2000 - Draper Laboratory

V 0 e jωt L R R R TL: (Z 0 ,β,I)C diaL diaThe impedance of the model in Figure 5 can be plotted as aSmith chart as shown in Figure 7. A 160-µm-diameterdiaphragm on a 295-µm-thick wafer is plotted. Resonanceoccurs when the plot crosses the left side of the horizontalaxis. Resonances at 1 and 3 MHz are marked.Figure 5. Lumped-element equivalent circuit model of monomorph pixel ona fluid-filled hole.representing radiation mass and resistance; [7] a transmissionline TL, which represents wave propagation down the fluidfilledholes; and lumped impedances C dia and L dia , which representthe compliance and mass of the air-backed diaphragm.This lumped-element model can be solved analytically, whichallows optimization to be carried out. In addition, the modelprovides insights that are absent from the finite-elementresults, although the FEA model is more accurate.Maximum sensitivity occurs at the system resonances whenthe reactances cancel, leaving only the radiation resistance.The transmission line provides an infinite series of resonancesapproximately when odd multiples of a quarter wavelength fitin the tube.The thickness of the wafer was chosen to provide resonanceat both 1 MHz and 3 MHz, the desired operating frequenciesof an underwater sonar. This calculated to a wafer thickness ofapproximately 295 µm, which was used for all fabrication.Figure 6 shows the average diaphragm deflection (per Painput) vs frequency for three wafer thicknesses, which give thefirst, second, or third resonance at 3 MHz. Achieving the firstresonance at 3 MHz required a wafer thickness of only 43 µm,which is too thin to handle. Achieving the third resonance at3 MHz required a 540-µm-thick wafer, but would not yield apeak near 1 MHz as desired. A fourth (diagonal) line is addedfor reference that shows the free particle displacement of a 1-Pa acoustic wave in water.Diaphragm Displacement (m)10 -1210 -1310 -14Figure 7. Smith chart showing diaphragm impedance in water from 0.1 to10 MHz. Resonances at 1 and 3 MHz are marked.FINITE-ELEMENT ANALYSISFEA was carried out on 13 different final designs. All pixelsmeasure 0.4 mm x 0.4 mm. The designs are grouped intothree general types: Quad designs (Figure 8) with fourdiaphragms per unit cell, racetrack designs with two or threediaphragms per unit cell (Figure 9), and Nona designs withnine diaphragms per unit cell (Figure 10).Figure 8 shows a quad cell with four diaphragms wired inseries. The concentric circular electrodes are apparent. Sincethe bending stress changes sign between the center and edgeof a diaphragm, only the outer portion of the diaphragm hasPZT and electrodes.The center small metalbond pad is forbump bonding to anelectronic "Transmit-Receive IntegratedCircuit" (TRIC) developedby LockheedMartin IRIS.10 -1510 5 10 6Frequency (Hz)10 7Figure 6. Average diaphragm deflection vs frequency for sensor diaphragmmounted on hole with three different lengths to achieve a resonanceat 3 MHz.Figure 8. Quad transducer has four diaphragms wired in series per unit cell(0.4 mm X 0.4 mm). Concentric electrodes and PZT layer areshown.Advanced MEMS Ferroelectric Ultrasound 2-Dimensional Arrays7

Infinite AbsorbingFluid ElementsCenterlineFluid-FilledHoleTransmissionLine LengthFigure 9. "Race-track" transducer has two elongated diaphragms wired inparallel per unit cell.DiaphragmFixed EdgeFigure 11. FEA problem as solved with ANSYS for radially symmetricdiaphragm (centerline shown) with diaphragm fixed at outeredge.QuadNonaRace TrackFigure 10. Quad, nona, and racetrack pixel designs, with 4, 9, and 2diaphragms per unit cell.ANSYS was used for all FEA computations. Figure 11 showsthe device simulated by ANSYS. The diaphragm model containsthe details of the various layers, and was electrically drivenby 1 V.The PZT response to the driving voltage deforms thediaphragm and generates an outgoing acoustic wave seen inFigure 12. The semi-infinite fluid-filled space is represented bya hemisphere of fluid elements ("mushroom cap"). The outerboundary of this hemisphere is made up of "infinite absorbingelements" that prevent reflections back toward the transducer.This technique works well as long as the hemisphere has aradius at least 1/3 λ.From the FEA model, pressure at distances of 0, 50, and 100µm from the mouth of the tube was tabulated as a function ofPressure (kPa)Figure 12. ANSYS FEA output showing diaphragm deflection and pressureisobars resulting from transducer electrical stimulation.121086423.2 MHz02.5 2.7 2.9 3.1 3.3 3.5Frequency (MHz)Figure 13. Quad 80 transmit response to 1-V sinusoidal input voltage.Pressure on axis at 3 distances (in µm) from mouth of hole.8Advanced MEMS Ferroelectric Ultrasound 2-Dimensional Arrays

Figure 16 shows a fabricated "race-track" design that has a linearlypolarized stripe of PZT down the center of thediaphragm. Figure 17 shows a cross section of an array thathas been cracked to reveal the hole side-walls. The etch wallsare close to vertical.Figure 15. Circular pixels are poled radially with concentric ring electrodes.POLING16 x 16 arrays were poled one row at a time using a probe cardwith 16 high-voltage electrodes and 1 ground electrode. Thepoling voltage was applied slowly using a ramp up, soak, andramp down of 2 min each. Large series resistors (40 MΩ) wereadded in series with each probe to prevent damaging the chipif any pixel contained a short circuit. Voltages up to 800 V wereapplied this way. Voltages between 800 and 1000 V resulted insparks through the air, but could be applied by covering thechip with several drops of Fluorinert FC-77 (a 3M dielectric fluid),which prevents air breakdown.Over-voltage caused cracking of the PZT, shown in Figure 18.The applied electric field causes a tensile strain of the PZT inthe direction perpendicular to the field equal to d 31 E. Whenthe tensile strain becomes too large (on the order of 0.3%), thePZT cracks. The cracks always form parallel to the electric field.Special breakdown test structures were included on the maskset for determining the maximum safe voltage as a function ofinter-electrode gap. Data on crack formation voltage for severaltest chips with 3-µm-thick sol-gel PZT is shown in Figure19. Although there is some experimental scatter, a field ofapproximately 20 V/µm was necessary to cause cracks to form;hence, the active devices could be poled at 10 V/µm with nocracks forming.TEST RESULTS RESULTSFigure 16. "Race-track" pixel as fabricated.The large 16 x 16 arrays were tested at Lockheed Martin bybump-bonding an electronics readout chip to the array. Inhousetesting was carried out by mounting small 3 x 3 arraysinto Kovar flat-packs with holes cut in the package to allowwater and acoustic signals to reach the chip (Figure 20). Thechips were wire-bonded to the flat-pack, which was solderedto a PC board, to which coax cables were attached. The entireassembly was then coated with silicone rubber for waterproofing,with a hole to allow sound to reach the sensor array.Figure 17. Through-wafer hole terminates on PZT monomorph diaphragm.Figure 18. Cracks formed in PZT by poling overvoltage. Cracks form parallelto the electric field.10Advanced MEMS Ferroelectric Ultrasound 2-Dimensional Arrays

Crack Formation Voltage (V)1200100080060040020000 10 20 30 40Inter-electrode Gap (µm)Figure 19. Crack formation voltages observed on test structures as functionof inter-electrode gap.Sensitivity (dB ref. 1 V/µPa)-190-200-210-220-230-240-250-260-270-280-2900 1 2 3 4 5 6Frequency (MHz)Figure 21. Receive data from several Mono-150 design pixels (notice broadpeak at 3 MHz).TVR in dB ref. 1 µPa/V at 1 m125120115110105100959085800 1 2 3 4 5 6Frequency (MHz)Figure 20. Kovar flat-pack used to mount the 3 x 3 test arrays.Figure 21 shows receive voltage sensitivity for several Mono-150 devices (single diaphragm, radius = 150 µm). Peakresponse is at 3 MHz, (the desired system frequency), withadditional peaks at 1.2 and 1.7 MHz. Noise floor of the systemis also shown. These data were compensated for 190 pF ofstray cable capacitance on each pixel.Figure 22 shows transmit response from several RT-115 (racetrackdesign, width = 115 µm) pixels. The transmit response isabout 115 dB// µPa/V at 1 m, which is very close to the predictedlevels for these devices. The "scalloping" or closelyspaced peaks and valleys in the frequency response are notyet understood, but may be related to a resonance in thepackaging.Figure 22. Transmit Voltage Response (TVR) from several RT-115 pixels on atest chip.CONCLUSIONSLUSIONSThis paper has described a new type of MEMS ferroelectrictransducer, the monomorph with in-plane polarization. Thisdevice allows sensitivity to be traded off against capacitanceto optimize the system signal-to-noise ratio. A dielectric layer(ZrO 2 ) compatible with PZT sol-gel deposition was developed.Arrays of MEMS sonar pixels were successfully built and tested.Sensitivity of -206 dB ref. 1V/µPa at 3 MHz was attained,which is an improvement of about 30 dB over previous MEMSferroelectric sensors at 0.8 MHz. [1] The devices are reciprocaland can be used for both transmit and receive.Advanced MEMS Ferroelectric Ultrasound 2-Dimensional Arrays 11

ACKNOWLEDGMENTS LEDGMENTSThe authors gratefully acknowledge support from the DARPASonoelectronics program as well as expert program monitoringfrom Bruce Johnson at NAVEODTECHDIV. I would also liketo acknowledge the contributions of M.C. Cardoso in chip fabricationand J.O. Miller in chip poling and testing.REFERENCESRENCES[1] Bernstein, J., S.L. Finberg, K. Houston, L.C. Niles, H.D. Chen,L.E. Cross, K.K. Li, and K. Udayakumar, "MicromachinedHigh Frequency Ferroelectric Sonar Transducers," IEEETransactions on Ultrasonics, Ferroelectrics and FrequencyControl, Vol. 44, No. 5, 1997, pp. 960-969.[2] Jin, X., I. Ladabaum and B.T. Khuri-Yakub, "The Microfabricationof Capacitive Ultrasonic Transducers," IEEEJournal of Microelectromechanical Systems, Vol.7,No.3,1998, pp. 295-302.[3] Swartz, R.G., Application of Polyvinylidene Fluoride toMonolithic Silicon PVF2 Transducer Arrays, PhD Dissertation,Stanford University, May 1979.[4] Mo, J.H., J.B. Fowlkes, A.L. Robinson, and P.L. Carson,"Crosstalk Reduction with a Micromachined DiaphragmStructure for Integrated Ultrasound Transducer Arrays,"IEEE Transactions Ultrasonics, Ferroelectrics and FrequencyControl, Vol. 38, 1992, pp. 48-53.[5] Chen, H.D., K.R. Udayakumar, C.J. Gaskey, L.E. Cross, J.J.Bernstein, and L.C. Niles, "Fabrication and ElectricalProperties of Lead Zirconate Titanate Thick Films," Journalof the American Ceramic Society, Vol. 79, No. 8, 1996, pp.2189-92.[6] Xu, B., Y. Ye, L.E. Cross, J. Bernstein, and R. Miller, "DielectricHysteresis from Transverse Electric Fields in LeadZirconate Titanate Thin Films," Applied Physics Letters,Vol.74, No. 23, 1999, pp. 3549-3551.[7] Kinsler, Frey, Coppens and Sanders, Fundamentals ofAcoustics,3 rd edition, John Wiley and Sons, 1982, pp. 191-193.[8] Beranek, Leo L., Acoustical Measurements, 1988 Edition,Acoustical Society of America/American Institute ofPhysics, 1988, pp. 113-122.[9] Bobber, R.J., Underwater Electro Acoustic Measurements,Peninsula Press, 1988, pp. 27-45.on BernsteinbiographiesJonathan J. Bernstein received a BSEE (Engineering-Physics) degree with honorsfrom Princeton University in 1977. At the University of California, Berkeley, he wasawarded MSEE and PhD degrees in 1980 and 1983, and received National ScienceFoundation and Hertz Foundation graduate fellowships. At Solavolt International,a Motorola subsidiary, he developed technology for silicon ribbon solarcell production by Chemical Vapor Deposition (CVD) of silicon and carbon films on reusableceramic substrates. He designed high-throughput parallel plate CVD reactors, and received theMotorola Engineering Excellence Award. At Draper, he is the Task Leader for micromechanicalacoustic sensors (hydrophones, microphone, and vibration sensors), accelerometers, andadvanced micromachined tuning-fork gyroscopes, in which capacity he has designed, analyzed,and fabricated these transducers. Dr. Bernstein is responsible for process development for siliconmonolithic sensors, including single-crystal silicon, polysilicon, PZT-on-Si, and electroformedmetal microstructures. He has also carried out process development to combine on-chipJunction Field Effect Transducer (JFET) circuitry with these micromechanical sensors. At Draper,he received the Distinguished Performance Award in 1992, the Best Invention Award in 1992 and1994, and the Draper Annual Award for Best Publication in 1993.12Advanced MEMS Ferroelectric Ultrasound 2-Dimensional Arrays: Biographies

ohn A Bottaribiographies biographiesbiographiesJohn A. Bottari is a Member of the Technical Staff in the Analog Systems Group atthe Charles Stark Draper Laboratory. Since joining Draper in 1999, his work hasincluded testing MEMS ultrasound transducers and designing the power systemarchitecture for a tactical mobile robot. He is currently working on the VHDLdesign of a filter used to reject mechanical resonance in the MK 6 gimbal controller.He received BS and MS degrees in Electrical Engineering from Tufts University in 1996and 1998, respectively, and is currently a member of the Institute of Electrical and ElectronicsEngineers (IEEE).enneth M. HoustonKenneth M. Houston is a Principal Member of the Technical Staff at the CharlesStark Draper Laboratory and Group Leader of the Analog Systems Group withinDraper’s Electronics Division. He has been with Draper since 1992, and hasworked on a wide range of projects related to acoustics, including testing micromechanicalultrasonic transducers, the concept design of an imaging sonarbased on these devices, and a system design and concept validation for a sonobuoy-basedmine-hunting sonar. Current projects include the design of a DSP-based electrolarynx systemwith the Massachusetts Eye and Ear Infirmary and the design of microphone/seismic sensor systemsfor vehicle detection. He holds an SB degree from Harvard University and a Master ofEngineering degree in Computer and Systems Engineering from Rensselaer PolytechnicInstitute. He is currently enrolled in the Master of Engineering Management program at theGordon Institute of Tufts University. He is a member of the IEEE and previously served for 2 yearsas chairman of the Providence Section.reg A. KirkosGreg A. Kirkos has been at Draper since 1995 and is currently a Senior Member ofthe Technical Staff. He has worked as an applications consultant for ParametricTechnology Corporation (now PTC), specializing in design-to-FEA systems forcompanies such as John Deere and Motorola. He is a Structural Analyst/Mechanical Design Engineer in the Mechanical Design/Analysis Group. His workinvolves mechanical design and analysis toward the development of many of Draper’s technologyareas, including MEMS, conventional inertial instruments, and robotic systems, includingDraper’s Vorticity-Control Unmanned Underwater Vehicle and WASP. He holds a BSME fromWorcester Polytechnic Institute.~---- -Advanced MEMS Ferroelectric Ultrasound 2-Dimensional Arrays: Biographies 13

aanan MillerRaanan Miller is a Senior Member of the Technical Staff at the Charles StarkDraper Laboratory where he is Task Leader for advanced MEMS accelerometerdevelopment. In addition to inertial sensors, he is also developing spectrometersfor chemical/biological agent detection. Dr. Miller received his BSc degree fromBoston University in 1990. He obtained MS and PhD degrees from the CaliforniaInstitute of Technology in 1992 and 1997.aomin XubiographiesBaomin Xu received his BEng degree in Inorganic Materials from Tsinghua University, Beijing,China, in 1986 and a PhD in Ceramic Science from the Shanghai Institute of Ceramics, ChineseAcademy of Sciences, in 1991. He is currently a Research Associate of Materials Science at theMaterials Research Laboratory (MRL) at Pennsylvania State University. His research interestsinclude ferroelectric thin and thick films and their applications in MEMS; piezoelectric materialsand devices for sensor, actuator, and transducer applications; and dielectric materials. Dr. Xu isa member of the American Ceramic Society and the Materials Research Society.aohong Ye14biographies biographiesYaohong Ye received her MEng degree in Electronic Materials from the University of ElectronicScience and Technology of China in 1995. Since 1998, she has been working at the MRL atPennsylvania State University as a visiting scholar. Her research interests include the preparationof ferroelectric films by sol-gel processing and their applications in MEMS.ric L. CrossEric L. Cross received his BSc (Hons) degree in Physics from Leeds University in 1948 and a PhDin Physics in 1953. He was a University Scholar, an Assistant Professor, and an ICI Fellow at theUniversity of Leeds. After a short period at the Electrical Research Association in LeatherheadSurrey, he moved to the United States to take up a position at the developing MRL atPennsylvania State University. For many years, he was Associate Director of MRL and was theDirector from 1985 to 1989. He is now an Evan Pugh Professor of Electrical Engineering atPennsylvania State University. His research interests are in dielectric and ferroelectric crystals;piezoelectric and electrostrictive ceramics; and composites for sensor, actuator, and transducerapplications and as components in "smart" materials and structures. Dr. Cross is member of theNational Academy of Engineering, a Fellow of the American Institute of Physics, the AmericanCeramic Society, and the American Optical Society. He is a United States representative for ferroelectricson IUPAP and a member of the Defense Sciences Research Council of ARPA.Advanced MEMS Ferroelectric Ultrasound 2-Dimensional Arrays: Biographies-- --

Gregg H. BartonSteven G. TragesserAutolanding Trajectory Designfor the X-34Based on the paper presented at the AIAA Atmospheric Flight Mechanics Conference and Exhibit, Portland, Oregon, August 9-11, 1999An Autolanding I-load Program (ALIP) is developed to designunpowered autolanding trajectories for the X-34 Mach 8 vehicle.The trajectory comprises geometric flight segments that are basedon the Shuttle approach and landing design (steep glideslope, circularflare, exponential flare to shallow glideslope). Enforcingphysical constraints such as loads, vertical descent rate, continuity,and smoothness reduces the design problem to a two-pointboundary value problem with conditions on the initial and finaldynamic pressure. Finding a solution required developing trajectorysimulation techniques that constrained the flight profile to aprescribed geometry. The design methodology can be extendedbeyond the autolanding flight regime by repeating the series ofgeometric segments and solving multiple two-point boundary valueproblems (one for each series). The techniques described in thispaper facilitate the rapid design of reference trajectories.ntroductionThe X-34 vehicle, depicted in Figure 1, is an 18,000-lb (drymass) vehicle that will make suborbital flights to test a variety ofnew technologies for reusable launch vehicles. The X-34 isdropped from an L-1011 and burns a single-stage liquid oxygen/keroseneengine. After reaching a maximum apogee of upto 250,000 ft at speeds up to Mach 8, the X-34 performs anunpowered descent to a horizontal landing. The vehicle is completelyautonomous; that is, no commands are uplinked to thevehicle throughout the entire flight. The primary contractor ofthe vehicle is Orbital Sciences Corporation. The descent guidancedevelopment, coding, and testing is subcontracted toDraper Laboratory.The X-34 was designed to behave similarly to the Shuttle in aneffort to minimize risk, cost, and development time. This designmaximizes the reusability of Shuttle software for the X-34. Thispaper deals with the automated design of the Shuttle’s approachand landing phase of the descent flight, which will be referred toas autolanding. The autolanding trajectory, discussed below, hasthe same segments as the Shuttle, and the guidance algorithmsare largely based on the Shuttle software.Figure 1.X-34 Mach 8 demonstrator.Autolanding Trajectory Design for the X-34 15

The motivation for developing ALIP presented in this paper isto: (1) decrease design time, and (2) eventually allow onboardtrajectory generation to add substantial robustness to offnominalconditions. In the first case, the algorithms allow analmost immediate redesign during X-34 development as thevehicle properties change or as new initial conditions are specified.This capability has obvious cost advantages, but it couldalso be used just prior to launch in order to redesign the referencesfor day-of-launch conditions (such as a measured windprofile). In the second case, as these algorithms are more fullydeveloped, they can eventually be put onboard the vehicle. Ifautomated trajectory generation can be performed in realtime onboard a reusable launch vehicle, then the vehiclecould more easily accommodate severe disturbances andabort situations.AUTOLANDING OVERVIEWThe unpowered autolanding trajectory begins at AutolandingInterface (ALI), which is typically at a 10,000-ft altitude, andends at touchdown on the runway. It comprises three segments,as shown in Figure 2: steep glideslope, circular flare,and exponential flare to shallow glideslope. [1] The steepglideslope maintains a constant flight path angle and is usedfor removing dispersions at transition. The next flight segment,the circular flare, employs a circular arc to linearlyincrease the flight path angle from the steep to the shallowglideslope. If the trajectory proceeded directly from the circularflare to the shallow glideslope, an acceleration discontinuitywould occur between these two flight segments.Therefore, an exponential flare segment is used in order todynamically smooth the transition onto the shallow glideslope.The vehicle flies the shallow glideslope (with a relativelysmall vertical velocity) until a final flare sets the craft on therunway. The X-34 differs from the Shuttle in that the vehicleflares to a terminal flight path angle to control the descentrate on touchdown. This maneuver will not be considered inthis paper, but is easily added on to the shallow glideslope aspart of the design process.The autolanding trajectory is completely defined by the parametersgiven in Table 1 (most of which are also shown inFigure 2). The mathematical description of the autolandinggeometry is given bywhere h STEEP ,h CIRC , and h EXP are the vehicle altitudes for eachof the flight phases at downrange distance, X. The value of Xis equal to zero at the runway threshold. The variable h SHAL isthe height of the shallow glideslope that the exponential flareasymptotically approaches and is given byh SHAL = (X - XAIMPT) tan γ 2 (2)(1)Figure 2. Autolanding geometry and phases.16Autolanding Trajectory Design for the X-34

Table 1. Autolanding trajectory design parameters.(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)γ1γ2XAIMPTHDECAYRXZEROX_KH_KXEXPHCLOOPσSteep glideslopeShallow glideslopeIntercept of the shallow glideslope with the groundAltitude above the shallow glideslope at exp. flare initiationRadius of curvature of the circular flareIntercept of the steep glideslope with the groundDownrange distance to the origin of the circular flame arcAltitude of the origin of the circular flame arcDownrange distance to exponential flare initiationAltitude for initiation of circular flareDecay rate of the exponential flarePROBLEM STATEMENTThe autolanding trajectory design problem is to find the valuesfor the parameters in Table 1 such that the vehicle landswith the desired dynamic pressure (or equivalently, velocity).TRAJECTORY DESIGN SOLUTIONThe parameters of Table 1 fall into three distinct categories.Parameters 1 through 5 are calculated only once and do notrequire trajectory simulation. Parameter 6 (XZERO) is used tomeet the constraint on the touchdown dynamic pressure.Parameters 7 through 11 are dependent on XZERO. Figure 3illustrates the procedure developed to solve for these parameters.An overview of the process is described directlybelow, and details on each of the tasks of Figure 3 are given inthe following subsections.First, variables are calculated or selected that do not requiredesign iteration. They are determined once and kept fixed.The variables γ 1 ,γ 2 , XAIMPT, HDECAY, and R are obtained basedon considerations such as robustness to uncertainties, loading,and vertical descent rate at touchdown. They are notdirectly related to the constraint on final dynamic pressure,and therefore do not need trajectory simulation to be determined.Next, a value for XZERO is guessed. This variable is the main"control variable" in the problem and will be used to satisfythe constraint on the dynamic pressure at touchdown. Theremaining variables (X_K, H_K, XEXP, HCLOOP, σ) are obtainedby imposing continuity and smoothness constraints on thetrajectory. These variables are all dependent on XZERO andmust be recalculated each time XZERO is changed.It is now necessary to find the dynamic pressure at touchdownfor the specified geometry and initial dynamic pressure(or velocity). Finding an analytic solution for the dynamicpressure is highly unlikely for the nonlinear (table-based)aerodynamic model. Therefore, a trajectory simulation is performed.Propagating the trajectories so they track the desiredgeometry was the most challenging aspect of the design andis discussed in detail below.Figure 3.YesTrajectory design procedure.The value of XZERO is varied until the touchdown dynamicpressure is within tolerance of the desired value. This constitutesa two-point boundary value problem where initial andfinal values for the dynamic pressure (or velocity) are specified.The details of the design procedure are described in the followingsubsections.VARIABLES NOT REQUIRING DESIGN ITERATIONThe steep glideslope, γ 1 , is an equilibrium solution of the vehicledynamics. The equations of motion for flight in the longitudinalplane of the vehicle, assuming a flat, nonrotating Earthand constant gravitational acceleration, g, are given byNo(3)Autolanding Trajectory Design for the X-34 17

(4)where the state variables v and γ are the velocity and flightpath angle, respectively, and the vehicle has mass, m. Theaerodynamic drag and lift are modeled asD = qSC D (5)L = qSC L (6)where S is the effective area, C D is the drag coefficient, C L is thelift coefficient, and q is the dynamic pressure, which is given bywhere ρ is the air density. The aerodynamic coefficients are afunction of Mach number, angle of attack, and control effectordeflection (elevon, body flap, rudder/speedbrake). The angleof attack is the control variable in the problem; the controleffector displacements are determined by trimming the vehicle(sum of moments is zero) at every integration step.The equations of motion are better behaved if we transformfrom velocity to dynamic pressure as one of the states.Dynamic pressure is a more slowly varying parameter for thisproblem because the monotonically decreasing contributionof velocity is partially offset by the monotonically increasingdensity. To transform Eq. (3), first differentiate Eq. (7)where ()’ denotes the derivative with respect to altitude, h, andρ’is assumed to be a known function of an exponential atmosphereor a numerically approximated quantity in the case oftabular data. Substituting Eq. (3) into Eq. (8) and using the relationship(7)(8)h = v sin γ (9)the equation governing dynamic pressure can be expressedas(10)dynamics will reject the error. Experience has shown that anonequilibrium flight path angle will not reject the largestpossible range of initial errors.The values for γ 2 , XAIMPT, HDECAY, and R were chosen basedon experience and Shuttle heritage. The design is not verysensitive to these parameters and a range of values shouldwork equally well as long as they are within certain bounds.The driver on the shallow glideslope is the vertical descentrate of the vehicle. If the vehicle is descending too rapidly, theloads at touchdown will exceed limits. However, the value ofthe shallow glideslope does not have to be determined accuratelybecause the final flare maneuver brings the vehicle tothe desired descent rate. The intercept point of the shallowglideslope, XAIMPT, will vary depending on runway lengthand rollout lengths. The value of HDECAY was selected empiricallyso that the exponential flare had a reasonable decayrate. The radius of curvature is arrived at by specifying anacceleration during the circular flare, NZ_FLARE. The curvatureis thenR = VCLOOP 2 / (g NZ_FLARE) (12)where VCLOOP is the velocity at HCLOOP. (Note that thedependence on VCLOOP indicates a weak dependence onXZERO. Thus, in order for the acceleration during the flare toexactly match NZ_FLARE, R should be included in the parametersbelow that are dependent on XZERO. However, in practice,the effect on VCLOOP is small so R can be fixed by thisinitial calculation.)VARIABLES DEPENDENT ON XZEROIn order for the trajectory to be feasible physically, it must becontinuous and smooth between the flight phases (i.e., nogaps or kinks). This requirement defines four constraints –position and velocity continuity for each of the phase transitions(steep glideslope to circular flare and circular to exponentialflare). These constraints can be obtained by settingthe altitude and altitude derivatives obtained from Eq. (12) ofthe respective flight phases equal. This results in four equationsthat can be solved for X_K, Y_K, HCLOOP, and XEXP. It iseasier, however, to solve these parameters using geometry,but the derivation is too lengthy to present here. The resultingsolution for the four parameters isThe equilibrium glideslope for a given q is obtained by solvingfor the flightpath angle and angle of attack that yieldsg = 0, q = 0 (11)for Eqs. (4) and (10). This requires the solution of two couplednonlinear algebraic equations. This solution is weakly dependenton altitude, so the steep glideslope is chosen based onthe equilibrium flight path angle at a particular point alongthe trajectory (generally, midway down the steep glideslope).This equilibrium solution is chosen for the steep glideslopebecause it provides the optimal flight dynamics for accommodatingoff-nominal initial conditions. That is, if the vehiclearrives at ALI with too little or too much energy, the natural(13)The purpose of the exponential flare is to eliminate any accelerationdiscontinuity between the circular flare and the shallowglideslope. Even with continuous altitude and altituderate references, a change in acceleration during the transitioncould cause transient behavior in the flight controller that is18Autolanding Trajectory Design for the X-34

undesirable so close to touchdown. Setting the radius of curvatureof the circular flare equal to the instantaneous radius ofcurvature at exponential flare initiation yields the final designparameter:The advantage of using altitude as the independent variableis now apparent since γ(h) and γ'(h) on the right-hand side ofEqs. (15) and (17) can be expressed in closed-form based onthe geometry of the trajectory. To find γ(h), we note(14)CONSTRAINED TRAJECTORY PROPAGATIONWith the geometry of the trajectory specified, it is now necessaryto propagate the vehicle dynamics in order to find thetouchdown dynamic pressure. This is not a standard initialvalue problem because the trajectory is constrained to theprescribed geometry and therefore implicitly dictates a certaincontrol history.One approach to simulating the trajectory is to use a guidancescheme (employing feedback control) to track the desiredgeometric profile. This is essentially the method used by theAutoland Shaping Processor [2] that predicts touchdown conditionsfor the Shuttle.The main disadvantage of using this techniquefor trajectory design is that the flight guidance andcontrol system adds unnecessary complexity to both thedesign and subsequent analysis. Also, this approach does notlend itself well as an onboard trajectory generation capability.This paper develops a different approach in which the geometricconstraint on the trajectory is substituted directly intothe equations of motion to reduce the order of the system.Walyus and Dalton [2] also used the known geometry to simulateautolanding. Their method uses curve-fit aerodynamicdata and propagates forward in the time-domain, which introducessome approximations that will not be present in thiswork. The extent to which these approximations reduce thepropagation accuracy (or require the use of more integrationsteps) has not been analyzed.We begin by changing the independent variable in the equationsof motion from time to altitude. This will greatly facilitatethe analysis since we are generally not concerned withthe time history of the trajectory. (The only situation in whichtime is important in the trajectory design is for abort trajectories.In this case, fuel is being dumped from the vehicle at aspecific rate, and the mass properties are a changing functionwith time. When dealing with this problem, it is possible toback out a time from the state variables.) Transforming theindependent variable in Eqs. (4) and (10) to altitude yields(15)(18)Differentiating Eq. (12) and substituting the result into Eq. (18)gives the flight path angle as a function of altitude for each ofthe flight phasesSteep, Shallow Glideslope:γ = γ 1 , γ = γ 2 (19)Circular Flare:Exponential Flare:(20)(21)The explicit dependence on X for the exponential flare couldbe expressed in terms of h if inversion of Eq. (12) were possible.However, the altitude equation for the exponential flare istranscendental in X, so X(h) is numerically solved using interpolationbetween values of h(x).To find γ’(h), we use chain rule differentiation of Eq. (18) to get(22)Differentiating the geometry in Eq. (12) and substituting theresults into Eq. (22) yieldsSteep, Shallow Glideslope:γ ’ = 0 (23)Circular Flare:Exponential Flare:(24)For convenience, Eq. (16) is solved for q(16)(17)(25)Substituting Eqs. (19)-(21) and (23)-(25) for the flight path andits derivative (corresponding to the particular flight phase)into the governing Eqs. (15) and (17) yields a system of theformAutolanding Trajectory Design for the X-34 19

q’ = f 1 (q, α, h) (26)q = f 2 (α, h) (27)Thus, enforcing the specified geometry has reduced the orderof the system so that the only state remaining is the dynamicpressure. Various numerical methods exist to solve the resultingdifferential-algebraic system given by Eqs.(26) and (27). [3],[4]The algorithm used for this application is shown in Figure 4.The outputs and equations associated with each task are givenadjacent to the diagram.The conditions for dynamic pressure and angle of attack at ALIare known, as depicted in Figure 4. The initial value of q isspecified by the previous flight phase, and the initial value ofα is determined from the solution for the steep glideslopedescribed in "Variables Not Requiring Design Iteration." Thederivative of q at the initial altitude is easily obtained from Eq.(26). To solve for the control and dynamic pressure at the nextintegration step, h+∆h, we first guess a value for α. (Note that∆h is negative for this problem.) This gives two estimates fordynamic pressure at altitude h+∆h, one based on the algebraicconstraint, q alg , and one based on the differential equation,q diff . The latter is obtained using an improved Euler numericalintegration technique (using derivative information at the currentand next integration steps). These estimates are comparedto determine if the equations are consistent, that is, ifthe proper control was chosen to stay on the specified geometry.The value of α(h+∆h) is varied by a secant root findingmethod until the dynamic pressure estimates match. Oncethe values converge, the solution for α(h+∆h) and q(h+∆h) iscomplete and the algorithm proceeds forward until touchdown.Figure 4. Trajectory propagation for a specified geometry.20Autolanding Trajectory Design for the X-34

This method proved to be very robust. If the profile underconsideration is not within the physical envelope of the vehicle,then no solution will be found for Eqs. (26) and (27). In thiscase, the search for an acceptable angle of attack exceeds thelimits of the aerodynamic tables.This method of finding the control for a desired profile can beextended to include a higher-fidelity gravity field and higherordereffects due to a spinning, oblate earth. A thrust profilecould also be included, but the additional degree of freedomand additional constraints may alter the manner in which theequations are solved.TWO-POINT BOUNDARY VALUE PROBLEM FOR XZEROWe return to the problem of finding the intercept value of thesteep glideslope, XZERO, so that the terminal constraint ondynamic pressure is satisfied. To solve this two-point boundaryvalue problem, we employ a shooting technique. [5] Foreach estimate of XZERO, a trajectory is propagated by the procedureoutlined in the preceding section. This simulationyields the dynamic pressure and control history along the candidatealtitude profile specified by XZERO. The value of XZE-RO is changed via a secant root finding method in order tosatisfy the constraint on the touchdown dynamic pressure.Varying XZERO essentially determines downtrack distanceflown during autolanding. Decreasing XZERO decreases thelength of the autolanding trajectory as shown in Figure 5. Thisresults in a shorter time of flight so that less energy is dissipatedand the vehicle lands with a higher dynamic pressure.Figure 5.RESULTSEffect of changing XZERO.To demonstrate the trajectory design algorithms, the DRM5mission of the X-34 is used. This mission reaches a maximumspeed of Mach 2.7 before beginning the unpowered descentguidance. At the autolanding interface of 10,000 ft, the vehiclewill nominally have a dynamic pressure of 330 psf.The required touchdown dynamic pressure is 123 psf (whichequates to a ground speed of 203 kn). Using ALIP to solve thetwo-point boundary value problem for this condition yieldsan XZERO of 7865 ft. The final trajectory design and thedynamic pressure are shown in Figure 6. The upper part of thetrajectory has been omitted to show more detail during theflare segments. Notice that the dynamic pressure stays fairlyconstant during the steep glideslope due to the use of theequilibrium solution. The dynamic pressure decays during thecircular and exponential flares to reach the desired value.Obtaining this solution requires 12 s on a 300-MHz Pentium ina nonoptimized MATLAB environment. For a poor initial guessof XZERO, the results may take two or three times longer.High-fidelity X-34 simulation software was used to validatethe solution and assess the performance. The 6-degree-offreedommodel includes a spinning, oblate earth, sensor models,actuator models, and actual X-34 guidance, navigation,and control flight software. The altitude and dynamic pressurehistories are indistinguishable from the plots in Figure 6.EXTENDING THE METHOD BEYOND AUTOLANDINGThe success of the algorithms in designing trajectories motivatedan investigation as to the applicability of the techniquesto higher altitudes. The flight phase prior to autolanding iscalled Terminal Area Energy Management (TAEM), whichbegins at about Mach 3.5 and ends at ALI (10,000 ft) for the X-34. To extend the design approach into TAEM, the pattern ofthe autolanding flight segments (steep glideslope, circularflare, exponential flare, shallow glideslope) was repeated priorto the autolanding interface. Note that the shallow glideslopeof this series of geometric segments coincides with and isequal to the steep glideslope of the autolanding flight phases.This approach is demonstrated in Figure 7 for an unpowereddrop test of the X-34. In one of the early designs for DRM4, thevehicle is dropped from an L-1011 at Mach 0.7 and an altitudeof 35,000 ft. A new set of flight segments has been computedfor the trajectory above ALI. This required the solution of twoseparate two-point boundary value problems – one for thedrop (above ALI) and one for autolanding. The terminal constrainton the dynamic pressure for the upper segment wassimply the initial value for the autolanding trajectory (seeFigure 7). In theory, this approach of designing a trajectorywith the geometric segments as building blocks can continuewell into TAEM.CONCLUSIONSThe method employed by ALIP for constructing a trajectory ofwell-defined geometric segments holds great promise forrapid trajectory design of an onboard reference for the flightguidance system. The trajectory design algorithms are automatedand converge on a solution very quickly. Using theAutolanding Trajectory Design for the X-34 21

4000Computed Altitude (ft)3000200010000-2.5 -2 -1.5 -1 -0.5 0 0.5x 10 4350300Qbar (psf)250200150100-2.5 -2 -1.5 -1 -0.5 0 0.5•Downrange (ft) x 10 4Figure 6. Trajectory design results: X-34 autolanding.4x 10 43Altitude (ft)210-12 -10 -8 -6 -4 -2 0 2x 10 4350Dynamic Pressure (psf)300250200150100-12 -10 -8 -6 -4 -2 0 2•Downrange (ft) x 10 4Figure 7. Trajectory design results: X-34 drop test.-22 Autolanding Trajectory Design for the X-34- --- -

slowly varying dynamic pressure as the dependent variableallows for large integration steps, thereby increasing computationalefficiency. If these concepts could be made sufficientlyefficient and robust, they can be implementedonboard atmospheric flight vehicles for real-time trajectorygeneration. Rather than rely on predetermined profiles, theonboard system would generate a profile based on the currentstate of the vehicle (as opposed to a fixed set of nominalconditions). In addition to increasing the performance andflexibility of the guidance system near nominal operatingconditions, this capability could be used to provide anautonomous abort capability.ACKNOWLEDGMENTSThe authors would like to acknowledge the X-34 guidance,navigation, and control team members Tim Osowski, MikeRuth, Dan Rovner, and Jennifer Henry of Orbital ScienceCorporation.REFERENCES[1] James, J., Entry Guidance Training Manual, NASA JohnsonSpace Center, Flight Training Branch, Houston, TX, July1988, pp 4-1.[2] Walyus, K.D. and C. Dalton, "Approach and LandingSimulator for Space Shuttle Orbiter Touchdown Conditions,"Journal of Spacecraft and Rockets, Vol. 28, No. 4,1991, pp. 478-485.[3] Betts, J.T., "Survey of Numerical Methods for TrajectoryOptimization," Journal of Guidance, Control, andDynamics, Vol. 21, No. 2, 1998, pp. 193-207.[4] Haug, E.J., Intermediate Dynamics, Prentice-Hall, 1997, pp.403-416.[5] Press,W.H., S.A.Teukolsky,W.T.Vetterling, and B.P. Flannery,Numerical Recipes in Fortran: The Art of ScientificComputing, Cambridge University Press, 2 nd ed., 1992. pp582-586.Autolanding Trajectory Design for the X-34 23

eg H. Bartonbiographies biographiesGreg H. Barton has been with Draper Laboratory since 1985. Currently, he is theChief Engineer for advanced guidance and control technologies for ReusableLaunch Vehicles (RLVs). This activity began in 1996 with the X-34 program andhas now grown to developing advanced algorithms for a new generation ofRLVs. His responsibilities over the years have included all levels of project developmentfrom concept design, algorithm and software development, to test and verification forflight certification. Management duties have included all levels from task lead, project lead, proposalmanager to program manager. In addition to these duties, he has served as technicalsupervisor and mentor for new staff and MIT graduates. He was responsible for managingDraper’s programs in Advanced Guidance and Control within the Space Programs Office during1999. These programs have spanned all aspects of Draper’s engineering disciplines includingadvanced algorithms, flight software, and hardware prototypes. Program management responsibilitiesinclude the Space Shuttle, Space Station, Mars Sample Return, and RLVs within NASAJSC, MSFC, and JPL centers, as well as commercial and international customers. In conjunction,Mr. Barton has planned and managed new business and program development.teve G. TragesserSteve G. Tragesser obtained a BS in Aeronautical and Astronautical Engineering from theUniversity of Illinois in 1992, an MS in Aeronautical and Astronautical Engineering from PurdueUniversity in 1994, and a PhD in 1997 with a dissertation on aerobraking with tethered spacecraft.Dr. Tragesser worked at Draper Laboratory for one and a half years developing descentguidance algorithms for RLVs. He is currently an Assistant Professor at the Air Force Institute ofTechnology. Areas of research interest include guidance of hypersonic vehicles; dynamics oftethered spacecraft; and trajectory modeling, design, and optimization.24Autolanding Trajectory Design for the X-34: Biographies-- ---

Robert HammettFault-Tolerant Input/Output (I/O) NetworksApplied to Ship ControlReprinted, with permission, from the Proceedings of the 12th Ship Control Systems Symposium held in the Hague, Netherlands, October 10-21, 1999Future ships will require sophisticated onboard control systems tocontrol machinery, automate tasks, optimize subsystem performance,and simplify maintenance. These controls will exploit theavailability of inexpensive computer processing, will use manysensors and actuation devices, and will provide for completelyintegrated and coordinated control of all subsystems. The crew'sincreased reliance on these automated functions makes it essentialthat they provide dependable operation despite equipmentfailure or battle damage, e.g., they must be fault tolerant anddamage survivable. An example of such a system is the U.S.Navy Seawolf submarine ship control. But the redundant sensorsand actuation systems used on Seawolf, with their associated electronicsand wiring, must be made more compact and affordablefor the approach to find widespread use on future ships. Thispaper describes how the use of data buses, intelligent sensors,and fault-masking actuation electronics can be used to constructInput/Output (I/O) networks that provide flexibility and growth,and that are highly dependable, affordable, and easily installed.These I/O networks can make widespread use of fault- and damage-tolerantsystems practical. The use of I/O networks complementsother efforts to make greater use of electrical actuationaboard ships. This paper explores the requirements for such systemsand examines some of the technology trade-offs that mustbe made, such as network media type (i.e., optical fiber, wired, orwireless), power distribution to network electronics, networktopology (ring or bus), network size vs speed, distributed vs centralizedI/O processing, and cross connections between I/O channels.The paper concludes by describing a concept for afault-tolerant network I/O system and discusses the steps neededto develop such systems for future ships.ntroductionShip Designers have already done much to place moderninformation systems aboard ships. The U.S. Navy Smartship andother projects have shown that it is both possible and beneficialto interface with and merge data from many of the ship's subsystems.[1] These high-level information networks make it possiblefor a small crew located in a central control room to monitorand control subsystems throughout the ship. But this alone doesnot make it possible to operate the ship with a small crew. Today,subsystems throughout the ship are manned to allow thesecrews to assume manual control and make repairs should battledamage or failures occur. Before these subsystems can be operatedunmanned, their dependability and damage survivabilitymust be improved substantially.The techniques for constructing highly dependable electroniccontrols have been established by applications such as fly-bywireaircraft flight controls. These same techniques were appliedsuccessfully to the U.S. Navy Seawolf submarine ship control. [2]The basic approach is to use physically redundant electronicsand a fault-tolerant digital controller with sophisticated softwareto detect faults and reconfigure to use the redundancy, allowingthe system to operate despite equipment failure or damage. Theproblem with widespread application of the approach used onSeawolf is cost and complexity. Modern ships already incorporatea myriad of wiring, junction boxes, electronic racks, and computersthat are difficult and costly to install and maintain.Duplicating, triplicating, or quadrupling this equipment to provideredundancy for fault tolerance is not practical.At least a partial solution to this problem can be found by theincreased use of data multiplexing and highly miniaturized electronicsplaced within intelligent sensors and actuators. Since thesensors and actuators are now part of a network of input andoutput devices, we refer to this arrangement as network I/O. Thedifference between a traditional system and a system that usesnetwork I/O is illustrated in Figure 1.Fault-Tolerant Input/Output (I/O) Networks Applied to Ship Control 25

ActuatorsSensorsIntelligent SensorsIntelligentActuatorsDedicatedWiringData Acquisition andActuator ControlElectronicsSharedBusWiringMiniaturizedDataAcquisitionElectronicsMiniaturizedActuatorControlElectronicsControlComputerControlComputerTypical Data Acquisitionand Actuator Control ElectronicsNetwork I/O Data Acquisitionand Actuator Control ElectronicsFigure 1. Network I/O approach reduces wiring and connections.As illustrated in Figure 1, using the network I/O approach, theanalog sensors and actuators that interface with data acquisitionand actuator control electronic units by dedicated wiringare replaced by "intelligent" sensors and actuators that communicateover a shared digital data bus. An obvious advantageof this approach is a reduction in the amount of wiringand connectors needed. For fault- and damage-tolerant systems,the reductions are increased by the number of redundantI/O devices needed. Less obvious but of equal or greaterimportance is the ability to program these intelligent I/Odevices to perform I/O processing functions such as signalprocessing (linearization, calibration, filtering, etc.), fault detectionand self tests, self identification, and actuator commandfault masking. The later of these, actuator commandfault masking, makes it possible to ensure that an actuatorsuch as a valve will operate correctly despite any single electronicsfailure.DEPENDABILITY REQUIRED OF THE I/O NETWORKA fundamental question when designing a system to be faultand damage tolerant is "what degree of dependability isneeded?" Designing a system to recover from a few obviousfailure conditions is a far easier problem than providing thelevel of dependability needed for life-critical applications,such as a fly-by-wire aircraft. Often, the way in which a systemis allowed to fail is of as great a concern as if it can fail. Forexample, failure of a weapon system to operate is of less concernthan premature detonation of that weapon aboard theship. So in this case, we would say that we require a very lowprobability of premature operation, but can tolerate a higherprobability that the system will fail to operate. Similarly, anunexpected shutdown of a turbine engine is less of a problemthan an explosive overspeed condition of the engine. In thiscase, we can tolerate the loss of function, but must design thesystem to avoid the overspeed malfunction. Yet other functionscan have catastrophic consequences only if they fail tooperate at the proper time, such as in response to an emergencycondition like a fire or flooding casualty. Another measureof system dependability is found in the rate at which falsealarm conditions are reported. Examples of false alarmsinclude a false indication to an operator that equipment isfaulty or an automatic reconfiguration to discontinue usingequipment that is not faulty. To summarize, some of the failuremodes that must be considered when designing a faultanddamage-tolerant system include:• Loss of function while operating.• Malfunction while operating.• Premature operation.• Failure to operate at the proper time.• Failure to cease operating at the proper time.• False alarms.• Degraded operation.A useful concept is that of categories of function criticality.Borrowing from the aircraft industry, three levels of criticalitycan be established: Critical, Essential, and Nonessential. A nonrigorousdefinition of each is as follows:• Critical: malfunction or loss of function endangersthe crew or the ship.• Essential: malfunction or loss of function requiresimmediate and effective action by the crewto prevent endangerment of the crew orship.• Nonessential: malfunction or loss of function will have nodirect effect on the safety of the crew or ship.It has been established by general practice that the only typeof system suitable for critical functions are those that useTriple Modular Redundancy (TMR). A TMR system (Figure 2)makes use of three redundant channels of control and relieson a mechanism to "vote" the output of these strings and tooperate the system based on a two-out-of-three consensus ofthe channels. Dual-redundant systems can be constructedthat provide protection against malfunction or prematureoperation, but are limited in their ability to protect against aloss of function while operating.26Fault-Tolerant Input/Output (I/O) Networks Applied to Ship Control

SUBSYSTEM CONTROL REQUIREMENTSRedundant SensorsIf A = B = C, Then X = A or B or CIf A = B ≠ C, Then X = A or B,If A ≠ B = C, Then X = B or C,A = C ≠ B, X = A or C,ABCno failureC is failedA is failedB is failedXBefore investigating how I/O networks can be applied to produceaffordable fault- and damage-tolerant subsystem controlsfor ships, it is important to define the types of systemsthat are to be controlled and to establish some basic I/Orequirements for these systems. Table 1 summarizes systemsthat are candidates for control or monitoring using I/O networksand lists a few of the basic control system functionsrequired. Tables 2 and 3 list sensors and actuators needed tosupport these functions. Typically, for control applications, thesensors and actuators are sampled/updated at rates of 1 to 10times per second.Figure 2.TMR systems provide highly dependable operation.Table 1. Subsystem control functions supported by network I/O.PropulsionElectrical Power Generation, Distribution, and ControlMechanical Fluid Systems• Hydraulic power• HVAC• Liquid cooling (chilled water, sea water)• High- and low-pressure compressed air• Steam• Machinery speed control• Pressure and temperature limiting• Torque sensing/limiting• Startup/shutdown sequencing• Performance monitoring (vibration, temperature, oil quality, etc.)• Voltage and frequency regulation• Load management/load shedding• Short circuit detection and recovery• Damage detection/reconfiguration• Performance monitoring (normal current loading, temperatures, vibration, etc.)• Electrical motor control• Pressure control and monitoring• Temperature control and monitoring• Flow control/diversion• Performance monitoring (temperatures, vibration, oil quality, flow rates, etc.)Fire Detection and SuppressionFlooding Detection and ControlWeapons Handling Machinery ControlMiscellaneous Winches, Lifts, Elevators, Cranes• Smoke, heat, flame detection• Suppression agent activation• Readiness testing• Seawater intrusion detection• Pump activation and flow routing• Readiness testing• Conveyor and lift control• Sequencing• Stores inventory measurement• Electric motor control• Sequencing• Performance monitoring (vibration, oil quality, actuation speeds, etc.)Fault-Tolerant Input/Output (I/O) Networks Applied to Ship Control 27

Table 2. Potential sensors requiring network I/O interfaces.• Pressure• Temperature• Shaft speeds• Linear and rotary positions• Limit switches• Flows• Liquid levels• Voltages• Currents• Switch positions• Lever positions• Vibration• Acoustic• Strain• Oil quality (conductivity, other)• Smoke• Heat• Flame (IR)• Seawater intrusion (conductivity)• Joysticks• Keypads• Barcodes, other ID devicesTable 3. Potential actuators requiring network I/O interfaces.Valve Control (modulated and two position)• Hydraulically actuated• Pneumatically actuated• Electromechanically actuated (EMA)Linear Actuator Control (modulated and two position)• Hydraulically actuated• Pneumatically actuated• EMAElectrical Switching and Contactors• Solid-state• ElectromechanicalElectric Motors• ac and dc• Fixed and variable speed, indicators, lamps, displaysNETWORK I/O DESIGN TRADE-OFFSThe design of an effective network I/O system requires theinvestigation and evaluation of a number of alternatives. Afew of these selected for discussion in this paper are:• Network media selection.• Powering the I/O network electronics.• Network topology.• Network data rate requirements.• Centralized or distributed I/O processing.• I/O channel cross-strapping.NETWORK MEDIA SELECTIONThree good choices are available for the network media: opticalfiber, copper wire, and wireless (RF or Infrared (IR)). Eachoffers advantages, and all have drawbacks. It is likely thatfuture ships will make use of all three, matching the strengthsof a particular media to a specific problem. The following arethe trade-offs between these media, with advantages precededby a plus (+) sign and disadvantages preceded by a minus(-) sign.Optical Fiber+ Supports very high data rates.+ Insensitive to Electromagnetic Interference (EMI).+ Does not propagate electrical faults (electrically isolated).- Optical signal is attenuated by connectors and splitters.- Attenuation makes fiber poorly suited to bus topologies(see I/O Network Topology).- Requires complex terminals.- Installation is complicated if wires are needed to power theterminal.Wire+ Small, low-cost terminals are available.+ Wiring can deliver power in addition to data.+ Electrical connections are easy to make.- Limited speed compared to optical fiber.- Requires shielding from EMI.Wireless (RF and IR)+ Very low cost to install (no wiring or fiber).+ Small, low-cost RF electronics are available.- To remain wireless, requires a battery (needs replacement)or power scavenging.- Vulnerable to EMI (RF).- Vulnerable to optical path obstruction by smoke, dust,equipment, or personnel (IR).- Limited bandwidth compared with wire or fiber.28Fault-Tolerant Input/Output (I/O) Networks Applied to Ship Control

POWERING THE I/O NETWORK ELECTRONICSThe sensors and actuation control devices that make up thenetwork require electrical power to operate. For sensors, poweris used to power the sensor and the terminal used to transmitits data. In the case of actuators, two types of power areneeded: power to operate low-power actuator control electronicsand the network data terminal, and power to operatethe actuator. Several options for delivering power to networkI/O terminals are:• Tap into the nearest source of ship’s power.• Distribute power as part of the data network.• Power the devices with batteries or other wireless sources.The strengths and weaknesses of each option are compared.As before, advantages are preceded by a plus (+) sign and disadvantagesare preceded by a minus (-) sign.Connect Network to Nearest Ship's Power+ Established practice.+ Unlimited power available.+ Power already needed by network I/O actuators.- Ship’s power can become a network I/O single point of failure.- If network I/O controls the power (a desired capability), failurescan cascade.- Wiring to ship’s power is expensive to install.- Ship’s power can transmit EMI or transient dropouts into thedata system.Dedicated Wiring as Part of I/O Network+ Combine installation of data and power wiring.+ Provides sensors with dedicated, uninterruptible power.+ Can power up sensors and actuation control for checkoutprior to applying main power.+ Allows actuator node to detect/report problems with actuationpower.- Adds wires or increases wire gauge, may constrain maximumsize of an I/O network.Battery+ Network devices are very easy to install.- Requires periodic battery replacement.- Limited power available.- Batteries may be large and temperature sensitive.- Many battery chemistries are explosive, corrosive, andexpensive.A preliminary conclusion, at least for sensors, is that includingpower distribution along with the data network offers thebest combination of ease of installation, network dependability,and system maintainability. This option becomes particularlyattractive if the same electrical conductors can be usedfor both the power and the data.I/O NETWORK TOPOLOGYSelecting an appropriate topology for the I/O network is animportant design consideration. Two viable topologies arethe ring and the bus. These two options are illustrated inFigure 3. Examined closely, the ring topology consists of aseries of dedicated, one-way data links from each terminal tothe next. Ultimately, the last terminal is linked to the first toform a ring. Data are sent from terminal to terminal by havingeach terminal rebroadcast any message that is not directed toit to the next terminal until the final destination is reached.The terminal that receives the message "absorbs" it; that is, itdoes not rebroadcast it. As an example, for terminal A to senda message to terminal B, it sends the message to terminal C onthe link from A to C, then terminal C rebroadcasts the messageto terminal B. A primary motivation for using a ring topologyis that each time the message is rebroadcast, its signalstrength is restored. This is ideal for fiber-optic media, whereeach physical terminal connection could otherwise attenuatethe signal until it became too weak to be received.Figure 3.Ring and bus topologies.RingTopologyBusTopologyIn contrast, the bus architecture uses a common physical link,the "bus," to connect all terminals to each other. Terminalsmust take turns using the bus to transmit messages directly toanother terminal(s). A subtle advantage of the bus over thering is that the bus can convey power as well as data to eachterminal on the bus. Table 4 makes a few comparisonsbetween the single-channel nonredundant ring and singlechannelbus topologies shown in Figure 3.Table 4 highlights that a number of single failures can disablethe entire network. For critical functions, this is unacceptable.To overcome vulnerability to single failures, redundancy isintroduced. For the ring topology, a proven approach is theuse of a redundant, counter-rotating ring and the addition ofa special switch that will bypass the terminal should the terminalfail or have its power interrupted (Figure 4). If a link isbroken, a terminal must detect this and reroute data throughFault-Tolerant Input/Output (I/O) Networks Applied to Ship Control 29

Table 4. Comparison of single-channel bus and ring topologies.Consequence of failure of a terminal to transmit Network disabled Loss of use of one terminal(or loss of power to the terminal)Uncontrolled broadcasts by a terminal (babbling) Network disabled Network disabledFailure of data path between terminals (opened or shorted) Network disabled All or part of the network disabledWorst-case number of transmissions to pass one message All terminals on the ring must transmit One terminal transmits(requires power)Number of optical or wire connections to each terminal Two optical fibers or wire pairs One wire pair connected to each terminalRing #1BypassBypassBypassBypassBypassBypassBypassDetails of Bypass Switch OperationNormalRing #2Figure 4. Dual counter-rotating ring network topology.the other ring to reach the destination terminal. The dual ringtherefore uses "active" reconfiguration to achieve fault-tolerance.If a single computer is controlling the terminal, the operationof the network is dependent on that computer, both tocorrectly reconfigure the network and to not malfunctionitself. As an example of a computer malfunction, the dual networkwould be disabled if the terminal were to transmit at thewrong time ("babble") on both rings.Similarly for the single-channel bus topology, allowing failureof the bus data path to interrupt all communications is unacceptable.Adding a second redundant bus can correct theproblem, but also requires "active" reconfiguration to determinewhen the backup bus should be used. So as for the dualring, if a single computer controls both terminals, both redundantbuses may be corrupted by a faulty computer, disablingthe entire network.As previously introduced, applications that require absolutedependability make use of TMR. Such systems have been constructedusing triple data buses for many years, e.g., the U.S.Space Shuttle and all fly-by-wire aircraft flight controls. Theuse of TMR with buses is illustrated in Figure 5.Note in Figure 5 that each bus terminal contains its own computerand that the system has been carefully segmented intoindependent channels. On the actuator side of the network, amajority voting device has been used to allow the actuator torespond to only a two-out-of-three consensus of bus commandmessages. Although implementation of the votingdevice will not be discussed, it can be constructed as a "passive"device that simply responds to the majority and does nothave single points of failure.Table 5 makes a summarized comparison of the single, dual,and TMR options. Single failures that lead to network failurehave been highlighted. Only TMR resists all single failures.30Fault-Tolerant Input/Output (I/O) Networks Applied to Ship Control

Redundant SensorsActuatorInternal Details of Bus Terminal (BT)Figure 5. TMR system using bus topology.Table 5. Comparison of simplex, dual, and TMR using rings and buses.Terminal Fails to Transmit Network Disabled Network Disabled Normal Operation After Normal Operation After Normal Operation WithActive Reconfiguration Active Reconfiguration Passive ReconfigurationUncontrolled Terminal Network Disabled Network Disabled Normal Operation After Normal Operation After Normal Operation WithTransmit Active Reconfiguration Active Reconfiguration Passive ReconfigurationMedia Failure Network Disabled All or Part of Normal Operation After Normal Operation After Normal Operation WithNetwork Disabled Active Reconfiguration Active Reconfiguration Passive ReconfigurationWrong Command Actuator Responds to Actuator Responds to Actuator Responds to Actuator Responds to Normal OperationisTransmitted Incorrect Command Incorrect Command Incorrect Command Incorrect Command Fault is MaskedNumber of Physical 2 1 4 2 3Connections RequiredNETWORK DATA RATE REQUIREMENTSThe speed required of the network depends on the number ofI/O devices on the network, the amount of data to or fromeach device, and the repetition rate at which it must be sent.Although modern data network electronics are capable ofhigh data rates at reasonable cost, it is still not practical todedicate a data transmitter and supporting electronics capableof gigabit-per-second rates to a single sensor. And as thedata rate increases, the length of data bus wiring must be limited,and more expensive and less durable cabling is needed.Fiber optics overcomes some of these disadvantages, butintroduces other difficulties as previously discussed.A key consideration is the total number of I/O devices on agiven network. A large ship might require thousands of I/Odevices, but this does not mean that each I/O network mustsupport very high data rates. Consider Figure 6, which illustratesthe use of a few large I/O networks compared with alarger number of smaller I/O networks.Ship’s Wide Area NetworkI/O NetworksA Few Large I/O NetworksIntelligent Sensors and ActuatorsMany Small I/O NetworksFigure 6. Alternative sizes of I/O network’s impact network speed required.Fault-Tolerant Input/Output (I/O) Networks Applied to Ship Control 31

Using many small I/O networks will directly reduce therequired data rate of each network. And there are many otherreasons to limit the size of the networks. One is to limit theextent of the effect of failure or battle damage to one smallnetwork. Another is that many types of network terminalsimpose limits on the number of devices that can be on onenetwork due to addressing limitations or electrical fan out.Also, if it is desired to provide power to the network terminalsover the same wiring as the data, the number of terminalspowered by the network wiring must be limited. By assumingan I/O network size of one hundred terminals per network,that each device transmits or receives 12 8-bit bytes of data,and that the average frequency of transmission is 10 times persecond, it follows that:Data rate = 100 devices x 12 bytes x 8 bits/byte x 10updates/s x 1.5 efficiency= 144,000 bits/sSince very low-cost terminals are available that can supportdata rates up to 1 Mb/s, smaller I/O networks can be built withreadily available and low-cost components. The conclusion isthat there is little motivation to use a very high data rate network.I/O PROCESSING – CENTRALIZED OR DISTRIBUTED?Since the I/O network introduces individual sensors and actuatorswith a network terminal containing at least rudimentaryprocessing capability, the question becomes how might thisprocessing be used to best advantage? Potentially, the processingcan be used to offload or even eliminate the need forcentralized I/O processing. In the extreme, distributed processingcould completely replace the need for any type ofcentralized processing by allowing "intelligent" sensors tocommunicate directly with intelligent actuators to control thesystem, but this discussion is limited to what I/O processingfunctions might be distributed. Table 6 lists traditional I/Oprocessing functions that might be distributed.As before, the trade-offs between distributed and centralizedI/O processing are listed:Centralized+ Established approach.+ Provides access to system-wide information to detect/isolatefaults, correct sensor outputs.+ Avoids interaction of complex software in multiple processors.+ I/O devices are kept simple, small, and lowest cost.- I/O design changes require central computer softwarechanges.- I/O repairs (replacement) can require central computer softwaredata changes.- Sending raw data rather than processed information consumesnetwork bandwidth.- Centralized closed-loop control can be problematic.Distributed+ High-level interface between sensor/actuator and centralprocessor (information, not data).+ Promotes the use and reuse of standard I/Ocomponents/processing functions (plug and play).+ Allows interchange of sensor/actuator units from differentvendors without software changes.+ I/O-specific processing and data contained within I/O unit(simplifies replacement).+ Sensor can preprocess raw data into compact or low-rateform for transmission.- Not practical to provide access to system-wide informationfor complete self-test.- Can introduce complex interaction between central anddistributed processors.- Can require reprogramming of each I/O device with application-specificlogic.The only conclusion to be drawn from this limited investigationis that there are many advantages to distributing I/O processing,but also limitations and concerns. For simplefunctions, such as linearization and filtering, these can be distributedeasily and offer many benefits. Distributing complexlogic, such as self-test and closed-loop control, requiresgreater caution during system design.I/O NETWORKS – CROSS CONNECTING DATA BETWEEN CHANNELSPreviously, it was shown how a well-designed TMR system ispartitioned into channels to prevent fault propagation. Thereare, however, reasons to consider providing data pathsbetween individual intelligent sensor and actuator terminals(see Figure 7).Table 6. List of I/O processing functions.• Sensor linearization• Sensor temperature and excitation level correction• Filtering and averaging (simple and Digital Signal Processor (DSP))• Individual sensor self-tests (range, rate, etc.)• Redundant sensor comparisons (self-test and sensor selection)• Calibration verification/correction• System-level fault isolation/reporting• Local actuator closed-loop control• Local actuator control sequencing• Network data error detection/recovery• Network data time-out detection• Local power monitoring• Equipment heath monitoring• Terminal self-identification• Terminal self-test32Fault-Tolerant Input/Output (I/O) Networks Applied to Ship Control

Redundant SensorsActuatorFigure 7. Adding cross connections between network terminals.The trade-offs are:System Without Cross Connections+ No potential for faults to propagate between channels(electrical or data).- Many two-fault conditions can lead to system failure.- Distributed I/O processing is limited to what can be performedwith only one channel of data.System with Cross Connections+ System can be made tolerant of most two-failure conditions.+ Adjacent channel data provides many options for distributedI/O processing.+ Adjacent channels can monitor each other's health, localizenetwork failures.- Adds hardware (size, weight, cost, power) to network I/Osystem.- Requires isolation to avoid fault propagation betweenchannels.It is believed that with careful design, both the added hardwareand the potential for fault propagation can be minimized.The benefits of having the cross connections outweighthese costs.CONCEPT FOR A FAULT-TOLERANT NETWORK I/O SYSTEMIn the previous discussions, it has been shown that many alternativedesign solutions exist for creating I/O networks. No singlesolution will be optimal for all applications. Yet there arecompelling reasons for pursuing a system design based on asingle approach, using interchangeable components with astandard interface. With the understanding that no singleapproach will be optimal and that technological advancescould render some of these conclusions obsolete, this sectiondescribes a concept for a fault-tolerant network I/O system.The following are the design requirements:• The network I/O system must be designed to support functionsthat are critical; that is, malfunction or loss of the functionendangers the crew or the ship. Providing this level ofdependability requires a TMR network I/O architecture.• The primary network media will be copper wire. Fiber-opticand wireless media will be used to a lesser extent to supportspecial installation requirements.• The network topology will be buses rather than rings. Aring topology complicates the installation of wiring andpower distribution, places too much dependence on thesurvival of individual network terminals, and makes eachterminal larger, more costly, and more power consuming.• The network will supply dedicated power to the I/O networkterminals over the same copper wire network used totransmit data.• The ship will use many small I/O networks. The size of thesenetworks will be limited to the maximum of several hundredterminals and a data rate of below 1 Mb/s.• The capability to cross-strap individual sensors and actuationterminals is provided.Figure 8 illustrates an overall system concept that meets theserequirements.CONCLUSIONSThe use of fault-tolerant I/O networks shows great promise forproviding highly automated and completely dependable controlsystems for ships. There appear to be few major technologicalbarriers to the development and introduction of thesenetworks. Many of the components needed are emergingfrom the development of industrial and automotive data multiplexing,data acquisition, and control systems. But theseapplications do not have the requirement for fault and damagetolerance, nor are the environmental requirements asdemanding. There is a need for a substantial engineeringdevelopment effort. Realizing the full benefits of theapproach will require highly integrated, miniaturized, andFault-Tolerant Input/Output (I/O) Networks Applied to Ship Control 33

Network Control ComputersTriple Twisted Copper Wire Bus NetworkSupplying Data and Terminal Power(Routed and Shielded for Survivability)NetworkTerminalPowerSuppliesSimplex Sensor or ActuationNode (Nonessential Function)Triple Redundant Sensor orActuator Node (Critical Function)Dual Redundant Sensor orActuation Node (Essential Function)Separate Source ofActuation Electrical PowerBus “Spurs” into Harsh orVulnerable AreasFigure 8. Concept for a complete network I/O system.harsh environment electronics, produced at moderate cost,that are not currently available "off the shelf." Commercial-offthe-shelfproducts will only appear after visionary developershave proved the concept with a successful "first-of-its-kind"system. This paper is intended to stimulate interest in such adevelopment.REFERENCES[1] Moschopoulos, J., "Advanced Integrated Control andInformation Systems in the U.S. Navy," Eleventh ShipControl Systems Symposium, Southampton, UK, 1997.[2] Hammett, R., "Seawolf Ship Control PerformanceMonitoring Provides Fault Tolerance and Simplified Maintenance,"Intelligent Ships Symposium II, Philadelphia,1996.34Fault-Tolerant Input/Output (I/O) Networks Applied to Ship Control

obert C. HammettbiographyRobert C. Hammett is a Principal Member of the Technical Staff at DraperLaboratory. He has degrees in both Mechanical and Electrical Engineering. Mr.Hammett has 22 years experience in the design, development, test, and manufactureof electronic control systems for aircraft, ships, and space applications. Hehelped pioneer the development of the first Full-Authority Electronic EngineControls (FADEC) for jet engines. These FADECs were first-of-a-kind systems that represented asignificant step forward in the use of digital electronic controls for life-critical applications. Hewas responsible for defining the fault detection, isolation, and recovery (FDIR) software thatmakes these FADECs fault tolerant and provides built-in-test features that enhance maintainability.Thesedesigns are in use today on aircraft such as the Boeing 757, 767, and 777. At Draper,he has developed fault-tolerant controls for several unmanned underwater vehicles, the Seawolfsubmarine, and the Kistler space launch vehicle. For the Seawolf, he was responsible for thedevelopment, test, and deployment of the performance-monitoring functions that detect faults,manage redundancy, and provide maintenance advisory information. The Seawolf was also aunique first-of-a-kind system, using a fault-tolerant computer and redundant I/O electronics toprovide a "swim-by-wire" system. Recently, Mr. Hammett has been the principal architect of theVehicle Health Management (VHM) software for the Kistler launch vehicle and a health managementsystem for a fault-tolerant unmanned underwater vehicle. He is currently the technicallead of a NASA program to develop I/O networks similar to those discussed in this paper forfuture space vehicles.Fault-Tolerant Input/Output (I/O) Networks Applied to Ship Control: Biography 35

Kaplesh KumarAnthony PetrovichTommy LeeMichael WattsRonald DennisHigh-Performance 18-GHz MicrowaveReadout Flexured Mass Accelerometer©1999 ION. Reprinted, with permission, from the Proceedings of the 55th Annual Meeting of the Institute of Navigation, June 28-30, 1999, Cambridge, MAA 10 8 dynamic-range flexured mass accelerometer usingmicrowave cavities and readout electronics operating at 18 GHzhas been shown to be capable of achieving 1-µg bias stabilitywith ±0.01˚C temperature control. Developments previouslyreported for the device operating at 10 GHz were implemented.The flexure cartridge, designed to eliminate anelastic and hystereticeffects during device static or tumble testing, was incorporatedinto the 18-GHz device. Two device configurations werefabricated and tested for resolution, drift stability, and tumblebias repeatability in a 1-g field. One was constructed entirely fromSuperinvar, save for the ceramic flexure cartridge, while the other,which also used a similar ceramic flexure cartridge, was of ahybrid construction employing Superinvar microwave cavities anda ceramic Zerodur housing. Two breadboard configurations oflow-noise 18-GHz electronics were designed, built, and tested successfullyto support these evaluations. The best results achievedincluded the following: resolution, 0.35 µg; drift stability with arigid insert substituted for the flexured proof mass,

Proof MassFlexure ConnectorCavityIncremental velocity is renderedHousingClampsCavity1 bit = 100 nano gSignal CouplersOpposing cavity sensors provide common-mode error subtraction•Figure 1. A schematic representation of the microwave FMA.Cavity CoverInvar•• @IIJ,~• ••••Top HousingProof MassFlexureCenter ShaftSpacer SleeveFlexureProof MassBottom HousingCavity CoverInvar or ZerodurInvarSapphireZerodurSapphireSapphireInvarInvar or ZerodurInvar•Figure 2. Assembly diagram for the FMA.38 High-Performance 18-GHz Microwave Readout Flexured Mass Accelerometer-- --- -

DESIGN CONSIDERATIONSThe overall goals of the FMA are shown in Table 1. Achieving1-µg bias stability in a 1-g field and maintaining

310518555310518550310518545310518540310518535Hz310518530310518525310518520310518515i3105185100 5 10 15 20 25x10 sFigure 3. FMA 18-GHz configuration with direct phase detection electronics.153785282153785280153785278153785276Hz153785274153785272153785270153785268i401537852660 2 3 4 5 6 7 8 9 10 11 12 13x10 sFigure 4. FMA 18-GHz configuration with IF electronics.High-Performance 18-GHz Microwave Readout Flexured Mass Accelerometer-- --- -

BIAS DRIFT STABILITYAn important requirement of the 18-GHz electronics and sensorreadout cavities is that they provide sufficient commonmodesubtraction to provide bias stability of

302728000302726000302724000302722000Figure 6. FMA 18-GHz configuration with direct phase detection, side B.HzHz3027200003027180003027160003027140000 100 200 300 400 500 600 700 800 900 1000x10 s3002001000-100-200-300-400x10 sFigure 7. FMA 18-GHz configuration direct phase detection, result of common-mode subtraction.42High-Performance 18-GHz Microwave Readout Flexured Mass Accelerometer

Hz29120600291205002912040029120300291202002912010029120000291199002911980029119700291196000 133 199 243 287 331 375 419 463 507 573 617 661 705 749•x10 sFigure 8. FMA 18-GHz configuration with IF phase detection, result of common-mode subtraction.525000000520000000515000000510000000Hz5050000005000000004950000004900000000 50 100 150 200 250 300 350 400•degFigure 9. Tumble with 0.020-in flexure cartridge and Superinvar housing.~---- -High-Performance 18-GHz Microwave Readout Flexured Mass Accelerometer 43

HzHz55000000548000005460000054400000542000005400000053800000536000000 5 10 15 20 25 30 35 40deg5160000051400000512000005100000050800000506000005040000050200000500000000 5 10 15 20 25 30 35 40iFigure 10. Tumble with 0.020-in flexure cartridge and ceramic housing.540000000deg535000000530000000525000000520000000Hz515000000510000000505000000500000000i44495000000 0 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55x10 sFigure 11. 18-GHz Superinvar housing -1-g to +1-g tumble showing repeatable and equal scale factors for the two sensor cavities.High-Performance 18-GHz Microwave Readout Flexured Mass Accelerometer-- --- -

In order to show bias repeatability, the response of the sensorwas displayed at one of the two orientations in a -1-g to +1-gtumble test. The two orientations are aligned to the gravityfield vector so that the response displayed is either the -1-g or+1-g state. Figure 12 shows the bias repeatability for theSuperinvar housing sensor. This figure shows only one frequencyoutput corresponding to the behavior sensed for theproof mass from one side of this two-sided output device. Fora 10-min period (60-s filter) involving 5 cycles between IA upand IA down, the rms variation was 2.64 µg. For an 8-min periodinvolving 4 cycles, the rms variation was reduced to 1.3 µg.Because one set of electronics used for reading out the sensorshad an excessive sensitivity to gravitational loading, thetwo-sided differential output for the Superinvar housing sensorwas degraded relative to that for the better side, whichshowed the 2.64 µg rms variation. The best rms variation for a10-min period (60-s filter) for the differential frequency outputof the two sides was 4.7 µg rms. The g-sensitivity for the suspectset of electronics was confirmed by switching the electronicsversus the cavities being probed and observing thatthe noisy signal followed the suspect electronics.The better side electronics provided similar bias repeatabilitywhen measurements were performed on the Zerodur housingdevice. An rms bias variation of 2.84 µg for one orientationwas measured for 10 min (5 cycles).SUMMARY(1) The 18-GHz sensor showed adequate resolution (

ACKNOWLEDGMENTSThe authors thank Mr. James Shearer, Program Manager, Capt.USAF (Ret.) Paul Lawrence (formerly of Air Force ResearchLaboratory – Wright Patterson AFB, Ohio), Draper Laboratory,and Maj. USAF Dennis Krepp, ICBM SPO - Hill AFB, Utah fortheir encouragement and support of this activity.REFERENCE[1] Petrovich, A., K. Kumar, T. Lee, and P. J. Lawrence, "RecentDevelopments in Flexured Mass Accelerometer Technology,"Proceedings of the IEEE 1996 Position Location andNavigation Symposium, Institute of Electrical andElectronics Engineers, Inc, New York, NY, 1996, (IEEECatalog Number: 96H35879), p. 19.aplesh KumarKaplesh Kumar received his B.Tech. in Metallurgical Engineering from the IndianInstitute of Technology-Kanpur, an MS in Metallurgy from the Stevens Institute ofTechnology, and an ScD in Materials Science from the Massachusetts Institute ofTechnology. He is a Principal Member of the Technical Staff at Draper Laboratoryand was Program Manager of the Flexured Mass Accelerometer effort. Dr. Kumarserves on the editorial board of International Materials Reviews and has numerous publications,including an Applied Physics Review monograph, and 14 U.S. patents.biographies46High-Performance 18-GHz Microwave Readout Flexured Mass Accelerometer: Biographies

nthony PetrovichAnthony Petrovich is a Principal Member of the Technical Staff at DraperLaboratory. He received a BS in Chemistry and Metallurgy in 1970, an MS in 1970,and an ScD in Materials Science in 1974, all from the Massachusetts Institute ofTechnology. Dr. Petrovich has participated in the design and development ofadvanced inertial instruments, and has led the development of the Flexured MassAccelerometer. He has authored numerous publications and reports in the areas of inertialinstruments, materials processing, and characterization of materials.om W. LeeTom W. Lee has a BSME from City University of New York and an MSME from theMassachusetts Institute of Technology. He has 25 years experience, primarily atDraper Laboratory, as a structural analyst, mechanical design engineer, sectionchief, virtual simulation group manager, and task leader on strategic guidancerelateddesign projects. He holds two patents on mechanical systems designedfor the aerospace composites industry.biographiesbiographiesHigh-Performance 18-GHz Microwave Readout Flexured Mass Accelerometer: Biographies 47

ichael R. WattsMichael R. Watts received his BSEE from Tufts University in 1996 and became aStaff Engineer at Draper Laboratory when this work was completed. He is currentlya Draper Fellow at the Massachusetts Institute of Technology. His primaryareas of research have been in the development of a heterodyne interferometerand a strategic Interferometric Fiber-Optic Gyroscope (IFOG).biographiesonald W. DennisbiographiesRonald W. Dennis received both BSEE and MSEE degrees from Wright State University in Dayton,OH. He is a Project Engineer in the Sensor ATR Technology Division of the Air Force ResearchLaboratory (AFRL/SNAR) at Wright-Patterson AFB, OH. Mr. Dennis has nearly 15 years experiencedeveloping navigational reference systems and sensors for strategic platforms.48High-Performance 18-GHz Microwave Readout Flexured Mass Accelerometer: Biographies-- ---- -

Ramses M. AgustinRami S. MangoubiRoger M. HainNeil J. AdamsRobust Failure Detection forReentry Vehicle Attitude Control Systems©1999 American Institute of Aeronautics and Astronautics. Reprinted, with permission, from the Journal of Guidance Control and Dynamics, Vol. 22, No. 6, November-December 1999This paper presents a robust failure detection methodology for theattitude control system of Reusable Launch Vehicles (RLVs). Inparticular, we consider the problem of estimating the thrust frommultiple jets firing from an RLV Reaction Control System (RCS), aswell as the related problem of distinguishing between failures inthe RCS and the aerosurfaces. For accurately known vehicle andsensor models, the Kalman filter provides the optimal estimate forthe jet thrust in the least-squares sense. During reentry, however,plant model uncertainties are a major problem for such a filter asthe vehicle’s aerodynamics vary widely with rapidly changingMach number, making gain scheduling impractical. Consequently,the Kalman filter's performance degrades. Even if the Mach numberwere known accurately, rapid gain scheduling may not bedesirable or even possible because of the large data storagerequirements it entails. Transient, robust H ∞ or game-theoreticfilters are proposed for next-generation RLVs, and a prototypedesign is demonstrated for the Space Shuttle Orbiter's attitudedetermination system. Simulation results demonstrate that therobust filters can be insensitive to plant model uncertainties overa much wider range of Mach numbers than a traditional Kalmanfilter, while remaining sensitive to failures in the aerosurfaces andthe RCS jets.ntroductionDuring reentry, RLVs typically rely on aerosurfaces and anRCS for attitude control. For example, the Space Shuttle Orbiter,used in this paper as a platform for designing a prototype robustfilter for next-generation RLVs, has aerosurfaces that consist ofthe rudder, elevons, speedbrake, and body flap, whereas the RCSconsists of bipropellant jets that are fired in the appropriatedirection to provide the desired thrust and to augment the aerosurfaces.During on-orbit operation, only the RCS is used. Shuttlereentry begins with the deorbit burn, where the Orbiter is orientedin a tail-first position and jets are fired to slow the vehicle andallow capture by the Earth’s gravity and atmosphere. The RCSjets reorient the Orbiter to a high-angle-of-attack, nose-first position.During the first part of reentry, the Orbiter is oriented to a40-deg angle of attack that is maintained until the vehicledescends and decelerates to below Mach 10. After passingthrough Mach 10, the Orbiter gradually reduces its angle ofattack to 10 deg. The RCSs are the sole attitude effectors until theatmospheric dynamic pressure is large enough for the aerosurfacesto become useful. At some altitude inside the transitionregion, the control surfaces are activated, and attitude control isprovided by both the RCS and the aerosurfaces. During the latestages of reentry, control is achieved using the aerosurfacesalone.Our objective is to detect and isolate failures in the attitude controlsystem. During reentry, the challenge is to distinguishbetween failures in the aerosurfaces (more specifically, theailerons, as rudders are not used in the early stages of reentry)and the RCS jets. In addition, we also would like to have an accurateestimate of the jet thrust, both in the presence and in theabsence of aileron failures. Correct isolation of a failure in theattitude control system is essential to obtaining accurate thrustestimates, which are used for monitoring the health of the RCS.For the purposes of this paper, we chose the reentry flight phasebecause it is the most difficult stage for the attitude control system’sfailure detection and isolation. Specifically, during reentry,the vehicle undergoes rapid changes in aerodynamic flightRobust Failure Detection for Reentry Vehicle Attitude Control Systems 49

properties as the Mach number decreases, and precise knowledgeof these flight properties is not available. The combinationof rapid change and model uncertainty also makes itdifficult to rely on a single, accurate Orbiter model and aKalman filter design based on that model, or on multiple modelsand gain scheduling. Moreover, the uncertainty also makesit difficult to distinguish between failures in the RCS and in theelevon control surfaces. Even in the absence of any failure, theKalman filter's jet thrust estimation performance degradesconsiderably in the presence of model uncertainty.Even if the Mach number were known accurately, rapid gainscheduling may not be desirable or even possible because ofthe large data storage requirements it entails. In general,however, modeling errors caused by inaccurate knowledge ofthe Mach number and other factors do exist. As a result, theKalman filter will not yield accurate thrust estimates. A desirablesolution, therefore, is a filter architecture that can distinguishrapidly between anomalies in the RCS and in theelevons. The filter architecture must also provide jet thrustestimates even if a jet misfires or an aerosurface fails.The most often used technique to add some measure ofrobustness to an estimator is to increase the design processnoise covariance, i.e., to overdesign the Kalman filter.However, there are limitations to this technique. In particular,it is shown in Refs. [1] and [2] that robust game theoretic or H ∞filters handle model uncertainties better than robustifiedKalman filters. The design of these filters is based on the smallgain theorem. [2]-[8] Such filters are robust to a general class ofnoise and plant model uncertainties. The steady-state robustfilter derived in Ref. [2] has been used in Ref. [9] to estimate theSpace Shuttle Orbiter’s RCS jet thrust. The results in Ref. [9]demonstrate that steady-state robust filters can yieldimproved thrust estimates for a wide range of Mach numbers,although the filter response is unacceptably slow.The effect of model uncertainty on the performance of stateestimators and failure detection algorithms is addressed inRefs. [2] and [10] through [13], among others. In Ref. [14], ageometric interpretation of the concept of analytical redundancyleads to a procedure involving singular-value decompositionsto determine redundancy relations that aremaximally insensitive to model uncertainties. An alternativeapproach is found in Ref. [15], where it is assumed that modelerrors may be deduced from the uncertainties of a set ofunderlying parameters. The partial derivatives of the residualswith respect to these parameters are then computed, and theresidual generator with lowest partial sensitivity is selected. InRefs. [16] through [18], a bound on the effect of model uncertaintieson the residual is estimated. This bound is then usedto set the threshold accordingly. Robust detection methodsbased on the unknown input observer include those of Refs.[12], [19], and [20]. Roughly speaking, an unknown inputobserver is a filter with an output of zero in the absence of failureregardless of the uncertainty and disturbance. Theauthors in Refs. [12] and [20] contributed significantly towardsolving the problem of robust detection using different techniques.Previous work using H ∞ techniques includes the work of Ref.[21], where a steady-state frequency domain-based filterdesign is used to attenuate the effect of disturbances. In Ref.[22], a steady-state H ∞ /µ robust filter for failure detection isintroduced. This filter requires the solution of two Riccatiequations and, as such, is robust to both disturbances andmodel uncertainty. As mentioned earlier, the use of transientfilters for this application is motivated by the work in Ref. [9].In this paper, we build on the work in Ref. [22] by designing afilter architecture based on the transient, discrete-time, robustgame theoretic or H ∞ filters derived in Refs. [2], [6], and [7] thatcan distinguish between failures in the RCS and the aerosurfaces.The failure detection architecture presented consists oftwo robust filters. One filter is tuned to estimate robustly thejet thrust even when jets misfire, whereas the other is tuned todetect failures in the aerosurfaces. Moreover, if a failure in theaerosurfaces is detected, then it is possible to reconfigure thejet thrust estimator to obtain an accurate estimate of the jetthrust. A one-filter architecture is also possible, but the twofilterarchitecture produced quicker detection.In the next section, we formulate the problem and demonstratethe deleterious effect of model uncertainty on the performanceof the Kalman filter, motivating the use of robustfiltering. In the Robust Estimation section, we present a briefformulation of the robust filtering problem. The robust filter’sequations and the Failure Detection and Isolation (FDI) architectureare also presented. Results and conclusions follow.PROBLEM DESCRIPTIONThe RCS consists of 44 bipropellant jets that, together with theaerosurfaces, provide attitude control and limited three-axismaneuvering capability. Thirty-eight of the jets are primaryjets, each capable of providing 870 lb of thrust in vacuum. Thejets are the sole attitude effectors for the initial part of reentrybecause there is insufficient dynamic pressure for the controlsurfaces to be effective in the thin atmosphere. As the vehicleloses altitude, falling into the increasingly denser atmosphere,dynamic pressure increases. The aerosurfaces are then activatedand augment the jets until there is sufficient dynamicpressure to control the attitude with the aerosurfaces alone.We consider the Space Shuttle Orbiter’s lateral dynamics forbank and sideslip. A linear model of the rigid-body reentryrotational dynamics is used as derived by Zacharias. [23]Although originally given in continuous time, the dynamicsare discretized for use with the Orbiter's digital computingsystem. The state-space representation for the Orbiter, linearizedat values of the state and control for an operatingpoint, is given byx k+1 =A k x k +B k u k +G k w k +Tθ ky k =C k x k +D k u k +E k v k (1)where x k is the state of the system at time k, u k is the aerosurface'scontrol input, θ k are the jet inputs, w k is the processnoise, y k is the measurement (angular body rates) vector, and50Robust Failure Detection for Reentry Vehicle Attitude Control Systems

v k is the sensor noise. In our case, there are four plant states.The first two states are the vehicle attitudes: bank angle andsideslip angle, and the next two states are the correspondingrates. The measurements consist of vehicle body roll and yawrates. These measurements are combinations of the bank andsideslip rate states, and the linearized equations at each flightregime relating these quantities are contained in the C kmatrix. As the Orbiter descends during reentry, the A k ,B k ,G k ,C k ,D k , and E k matrices change in accordance with changes inMach number and angle of attack. In addition, a new state isdefined for the jet thrust estimate. A simple, scalar, high-bandwidthGauss-Markov model is used to represent a multiple-jetfiring in a particular direction, i.e.,θ k+1 = a θ θ k +g θ v k (2)The parameters a θ and g θ determine the bandwidth andamplitude of the model. For simplicity, we consider jets firingin only a single direction. Multiple Gauss-Markov models canbe added to estimate thrust in multiple directions. The precedingmodel is augmented to the plant model of Eq. (1) todesign the Kalman and robust filters. The interaction betweenthe jets and the states is also modeled by augmenting thematrices A k in the same equation. Consequently, the actual filterdesign contains five states. Only a single additional state isnecessary to model multiple jets firing in the same direction.Multiple jets create scalar increases in thrust, easily modeledby a first-order Gauss-Markov process. This state, after beinginitialized to the number of jets commanded to fire, estimatesthe total thrust in a specific direction. Note that this additionalstate is not able to identify which jet has failed, but allows3000the comparison of the total number of jets firing in a specificdirection with the total number commanded to fire, enablinga failed jet to be detected.Our objective is to design a filter based on a time-invariantplant model, obtained by linearizing around a certain operatingpoint, that performs well over a flight envelope or a rangeof operating points. This makes gain scheduling at each timestep unnecessary and provides robustness to uncertainties.As mentioned earlier, the filter is to robustly detect and isolatefailures in the lateral dynamics attitude control system, i.e., todetermine whether a failure is in the elevons or in the RCS. Anadditional goal is to obtain a robust estimate of the jet thrust,even if a jet misfires or a failure in the aerosurfaces occurs.Correct isolation between jet and aerosurface failure is essentialfor accurate jet thrust estimation because without properfailure isolation, irregularities in thrust due to aerosurface failurecould be interpreted as off-nominal jet performance.To motivate the need for robust filtering, a Kalman filter isdesigned based on the linear, time-invariant state-space modelfor the nominal condition of M = 7.5 and an angle of attackof α = 35 deg. This filter is then tested on both the designplant and on a perturbed plant model based on a Mach numberof M = 8.8 and α = 38 deg. A step-on-step-off commandwith two jets firing in the same direction is used in this simplesimulation. Aerosurfaces are used for trim adjustment inresponse to fluctuations in atmospheric conditions as modeledby the process noise.Simulation results for the Kalman filter are shown in Figure 1.The solid line represents the commanded jet thrust magnitude.The dashed lines represent a 15% error margin for the25002000Thrust (lb)150010005000-50011 11.5 12 12.5 13 13.5 14 14.5 15Time (s)Figure 1. Kalman filter’s jet thrust estimates for a nominal and perturbed plant.Robust Failure Detection for Reentry Vehicle Attitude Control Systems 51

thrust estimate. The jet thrust estimates given by the Kalmanfilter are represented by the dash-dotted line for the designplant and a dotted line for the perturbed plant.The estimates for the nominal plant and for the perturbedplant reveal that the Kalman filter works well for the correctplant model, but provides a severely degraded estimate whenthe model is perturbed. Note that the error in the degradedestimate for the perturbed plant is large enough to lead to theconclusion that more jets are fired than commanded. As mentionedin the introduction, it is desirable to avoid, or at leastminimize, gain scheduling. Gain scheduling a Kalman filterwould be necessary because of the presence of model uncertaintyand the rapid change in aerodynamic conditions duringreentry, which, in the mid-reentry region of concern, movefrom Mach 10 at about a 170,000-ft altitude to less than Mach2 at about a 75,000-ft altitude in approximately 20 min. Newlinear environment models that would need to accompany aset of gain-scheduled traditional Kalman filters would be necessaryat an inordinately large number of linearization pointsthroughout reentry because of the highly nonlinear nature ofthe flight regime during the reentry flight phase. The effectivenessof the Kalman filter would be reduced because of thetime lost to multiple reinitializations of the filter and increasedoverhead in switching linearization points. In comparison, thetime lost to the slight overhead of robust filters with a muchless restrictive number of gain-scheduled linearization pointsis minimal.The next section presents a general formulation of the robustestimation problem and the robust filter equations. Theseequations are derived in Ref. [2]. The section also describes thefilter architecture used for our robust failure detection andthrust estimation problem.ROBUST ESTIMATIONFigure 2 is a general input/output representation of a nominalplant P with modeling uncertainties ∆ and an estimator F. Thevector r represents the combined process and measurementnoise, x 0 the initial state vector of the nominal plant,x 0 the initialstate estimate, y the measurement vector, and e the estimationerror. A known control input signal u may beaccommodated easily in this formulation as part of the exogenoussignal r (see Ref. [2] for details). The signals ε and η representthe interaction between the nominal plant and theperturbation. In what follows, ||r|| represents the 2-norm ofthe signal r over the time interval of interest. For instance, indiscrete-time ||r|| ´ (Σr i ’r i ) 1/2 is the l 2 norm of r. The norms ofthe other signals, namely η, ε, y, and e, are defined similarly.Weights can be added to these models, or alternatively, theeffect of the weights can be incorporated by varying the plantmodel parameters (to be described shortly). Finally, ||x 0 || 2 w 0and||x 0 – x 0 || 2 represent the weighted Euclidean norms of the initialcondition x 0 and the initial estimation error x 0 –x 0 ,P -10respectively.The robust estimator seeks to bound the induced l 2 norm of theoperator from the input disturbances r and initial estimation errorrx 0η(x 0 – x 0 ) to the estimation error e, provided the model perturbations∆ have bounded induced l 2 norm. Mathematically, thistranslates to the following performance criterion:(3)The approach used to achieve the performance goal of Eq. (3)consists of treating η and x 0 as additional inputs to P andtreating ε as an additional plant output. The revised objectiveis then to bound the induced norm of the mapping from theaugmented input to the augmented output by 1. Thus, a newperformance criterion is defined aswhereEquation (4) is a robust performance or small gain condition.It is easy to show that satisfying this criterion guarantees theoriginal performance condition of Eq. (3). It is in turn possibleto achieve the condition of Eq. (4) by solving a minmax estimationproblem with the objective function given bywhereFigure 2.subject to the plant’s dynamic constraints. Specifically, if asolution to the preceding minmax problem is possible for avalue of γ≤1, then J 3 – γ 2 ||x 0 || 2 w 0< 0, which is equivalent to satisfyingEq. (4). The original criterion of Eq. (3) is then satisfiedas well.εyx0General representation of the robust estimation problem.e(4)(5)52Robust Failure Detection for Reentry Vehicle Attitude Control Systems

The preceding performance criterion is different from that ofthe Kalman filter’s. Specifically, the robust filter attempts tominimize the ratio of the signal 2-norm of the augmentedoutput to the signal 2-norm of the augmented input, whilethe Kalman filter minimizes the 2-norm of the estimation errorsquared. Furthermore, because of the relationship of thegame-theoretic optimization to risk sensitivity, the differencebetween the Kalman filter and the robust filter has a stochasticinterpretation as well (see Refs. [2] and [7] for details).ROBUST DISCRETE FILTER EQUATIONSWith r k = [w ’ kv ’ kυ ’ ku ’ k] ’ and d k = [r ’ kη ’ k] ’ , we write a time-varyingnominal plant state-space representation asx k+1 =A k x k + B k D k(6a)ε k =S k x k + T k d k(6b)e k =M k (x k – x k )(6c)y k =C k x k + D k d k(6d)The matrices S k and T k represent how the state and disturbanceinteract with the perturbation ∆ and are a function ofthe uncertainties in A k ,B k ,C k , and D k (see for instance Ref. [2],pp. 61, 62). The error e k is defined by a weighting matrix M kthat may be used to emphasize certain states over otherstates. We note that the state x k in Eqs. (6a) through (6d)would include plant states, as well as any additional states,such as the thrust θ k in the preceding section, and possibleshaping filter states for modeling disturbance processes.The solution for the robust estimation problem of Eq. (5), asderived in Ref. [2], is then given aswhere(7a)(7b)(8)(9)(10)(11)(12)(13)(14)and the matrices X k and P k are, respectively, positive definitesolutions to the following two Riccati equations:(15)(16)In the preceding equations, the assumption is made that Z kand H k are positive definite and that X 0< γ 2 X 0.The solution to this problem is an extension of both the H ∞optimal estimator and the Kalman filter for nominal systems.If there are no model perturbations, then S k = T k = 0 in Eq. (6b),so that the Riccati Eq. (15) for X k is superfluous, i.e., X k = 0. Inthat case X 0 = 0 by assumption. The estimator is reduced tosolving one Riccati equation based on the nominal plantdynamics. The Riccati Eq. (16) and the gain (Eq. (7b)) are thenthe same as those of the Kalman filter, except that the term-1H k= (P k– γ -2 M ’ M ) replaces P -1 k k k. In the case of the Kalman filter,P -1 kis a measure of the information available at time k priorto taking a measurement. Subtracting γ -2 M ’ M from P -1 k k kmeansthat we believe we have less information available. This filteris the standard game theoretic filter. Note that the smaller thechoice of γ, the more conservative we are.With ∆ ´ 0,γ is no longer constrained to be less than unity. Theminimum value of γ for which H k > 0, A k gives the solution tothe optimal H ∞ problem. On the other hand, if γ å ∞, thenH k-1 å Pk in both the Riccati equation and the optimal gain. Inthat case, using any M k , one recovers the Kalman filter. When∆ = 0, the minmax or game theoretic estimator can, therefore,be viewed as an extension of the Kalman filter and the H ∞optimal estimator, where decreasing the design parameter γtrades off-nominal performance in the minimum variancesense to provide robustness to the disturbance modelingerror.ROBUST FDI FILTER ARCHITECTUREFigure 3 shows the filter architecture used for attitude controlsystem fault detection. The output from the plant goes intotwo robust filters. One of the strengths of robust filter designis that various ways of representing the uncertainty andweighting matrices are possible. This flexibility can be used totune in each of the two filters to a different set of states. Thiscan be done, for example, by varying the values of the weightingmatrix M k in Eq. (6c). Other techniques are also possible,such as those involving the matrices S k and T k in the precedingsection. For more details, see Refs. [2]-[3] and [5].In the architecture of Figure 3, one filter concentrates on theRCS, whereas the other concentrates on the aerosurfaces.Filter F 1 provides accurate thrust estimates. The filter isdesigned to robustly minimize the sum of squared errorΣ(θ k – œ k ) 2 (Eq. (2)). Specifically, the filter parameters are selectedso that the induced 2-norm of the mapping between thedisturbances as well as the possible aerosurface failuresRobust Failure Detection for Reentry Vehicle Attitude Control Systems 53

x 0ηεryx 0Figure 3.Filter architecture.x 0(modeled as potential inputs) on the one hand, and the estimationerror on the other, are kept to a minimum over theentire range of plants under consideration, i.e., Mach 5.5 to 8.8.Over the same range, the mapping between a jet thrust inputand the thrust estimate is kept as close to unity as possible.The second robust filter F 2 is designed to detect failures in theaerosurfaces only, and is therefore insensitive to perturbationsin the jet thrust input. This filter is designed to minimize theoutput residual square error Σ(y k – y k ) 2 . In this design, parametersare chosen so that the induced 2-norm of the mappingfrom disturbances and thrust input to output residual error isas small as possible over the entire range of plants. On theother hand, the same mapping is highly responsive to aerosurfacefailures, modeled as inputs.The details of the two filters’design can be seen in Ref. [25]. An alternativedesign,where an aileron failure state is introduced usinga Gauss-Markov model, is also possible (see Ref. [2]). In that case,the objective would be to minimize that state’s estimation error.There is no systematic design procedure that can lead easilyto the desired robust filter design. Any robust filter design isan iterative procedure that requires the analysis of differentmappings.The detection logic works as follows. Once an aileron failure isdetected by the second filter, the first filter can be modifiedeasily to provide accurate thrust estimates. This is done byadjusting the parameters so that the induced 2-norm of themapping between the newly discovered failure and the thrustestimation error is kept as low as possible. If a jet failureoccurs, filter F 1 can provide an accurate thrust estimate, thusallowing for jet failure detection.Because we do not consider simultaneous failures for the purposesof this paper, using these two filters in parallel allows forthe proper isolation of both kinds of failures in the SpaceShuttle Orbiter’s attitude control system, aerosurfaces, andjets, and permits accurate thrust estimates even in the presenceof aerosurface failures. Finally, a one-filter architecture ispossible, but the detection response is slower.A primary strength of robust filters, lessened performancesensitivity to specific off-nominal conditions, also contributesto slower estimation time response when compared with ae 2ThrustEstimateResidualsDecisiontraditional Kalman filter designed around the appropriateplant dynamics. Compare, for example, settling time afterdetecting jet thrust in a traditional Kalman filter design, Figure1, and a robust estimator, Figure 4, in the next section. Onemethod used in this robust FDI filter architecture to speed upthe robust filter’s response time is to reset periodically theRiccati matrix P to larger values. This causes the robust estimatorto converge more quickly to a perturbed estimate thanit otherwise would with mature, small values in the Riccatimatrix. Because an off-nominal condition is by its nature atransient phenomenon, past information contained in theRiccati matrix is not useful, so little is lost by resetting thematrix in this manner. Naturally, a careful trade-off must bemade between quicker response times to perturbations withmore frequent Riccati matrix resets and decreased estimationaccuracy while the reset filter converges to a new solution.In the next section, we will demonstrate the use of this resettingtechnique for the robust filter and compare results withthe same technique applied to the traditional Kalman filter. Incases where we have applied this technique, we will refer tothe filter as a transient filter because resetting the Riccatimatrix will cause the gains to be time-varying. Where we alsoexamine the performance of the robust and Kalman filterswithout this technique, we will refer to the filter as a steadystatefilter because after initialization, the gains will have convergedto a steady-state value and will effectively no longerbe time-varying.RESULTSThe robust FDI architecture just described is tested using thesame simulation used to present the Kalman filter results inthe Problem Description section. Nominal and perturbed systemswere defined based on different flight regimes.Specifically, parametric perturbations are derived from twosets of linear time-invariant system matrices for different flightoperating conditions (see Ref. [2], pp. 60-62), namely, M = 7.5, α= 35 deg and M = 8.8, α = 38 deg, where M is the Mach numberand α is the angle of attack. Certain nonlinear effects, suchas jet self-impingement, variation in atmospheric pressure,and other transient effects are ignored. The operating pointsM = 7.5 and 8.8 were chosen because of the high degree of jetactivity present between these two Mach numbers duringreentry. Whereas good performance with the robust filterswill be demonstrated at these two points, it is anticipated thatsome gain scheduling of robust estimators may still be necessaryover the entire reentry flight regime. Nonetheless, gainschedulingrequirements for robust estimators will besignificantly less severe than gain-scheduling requirementsfor traditional Kalman filters.Referring back to Eqs. (1), the process noise input matrix G k istaken as 1/20 × (diagonal elements of B) to indicate a ±5%uncertainty in aerosurface deflections. A sampling interval ofT = 0.005 s is chosen to achieve the minimum average estimationerror in shortest convergence time. Whereas this timeinterval is shorter than the current Orbiter sample interval, it isrealistic for the next-generation RLV for which this FDI architectureis designed.54Robust Failure Detection for Reentry Vehicle Attitude Control Systems

JET THRUST ESTIMATIONWe will first discuss results for jet thrust estimation, which isgiven by filter F 1 in the architecture of Figure 3. Simulationresults comparing the performance of the Kalman and robustfilters are shown in Figure 1 and Figure 4. The filters arerestarted at the start and end of the jet firing sequence, i.e.,whenever the number of jets commanded changes. Theseplots show the effectiveness of the robust filter’s insensitivityto model uncertainty. The dashed lines indicate the 15% errormargin, and it is clear that the robust filter estimates are withinthat margin for both nominal (Mach 7.5) and perturbed(Mach 8.8) plants, whereas the Kalman filter estimates lie outsidethe error margin for the perturbed plant. Note also thatthe performance of the robust filter in Figure 4 for either thenominal or perturbed plant compares well with that of thenominal, or optimal, Kalman filter shown in Figure 1.Comparable performance of both filters also indicates thatthe filters have similar bandwidths. Although the robust filterdoes, in fact, have a smaller bandwidth than the Kalman filter,this is not always the case. Generally speaking, the robust filteris designed to have a low gain in the region where uncertaintyin the model dynamics has the greatest negativeimpact (see Ref. [22] for an example). Where it is possible touse this reduced gain approach in robust estimator design,the quick response time of the Kalman filter is preserved asmuch as possible. The ability to design the robust filter in thismanner for this RLV application is aided by accommodatingsome gain scheduling of filters over the widely varyingdynamics of the various reentry regions.In this simulation, jets were fired during three intervals over a15-s simulation period. Figure 5 shows the robust filter’s transientperformance during the first 2 s of that simulation period.One of four jets commanded fails to fire at 0.35 s andremains off until 0.74 s. The spikes in the plots of Figure 5 arecaused by the filter’s restarting. Overall, these results demonstrateclearly that while the transient robust filter is insensitiveto model perturbation, it is highly sensitive to unexpected jetmalfunction. Note also that the performance of the robust filteris maintained not only at the two flight regimes that wereused in the filter’s design, Mach 7.5 and 8.8, but also at anintermediate regime that was not used explicitly in the filter’sdesign. This demonstrates that the robust filter is usable overa range of flight regimes and is not just tuned specifically forfavorable performance at only two points in the regime.Results comparing the Kalman filter with the robust filter overthe entire 15-s simulation period are presented in Table 1. Twodifferent simulation runs are used: one where jets fire as commandedand the other with jet failures. The firing patternscommanded are the same for both simulations and are shownin part in Figures 1 and 5. The sum of squared error is the performancemeasure used for comparison. The term unfailed jetin the table refers to a pattern without jet failure, whereas theterm failed jet refers to a firing pattern with jet failures, such asthe one shown in Figure 5. The steady-state filters are resetonly when new jets are fired, whereas the transient filters arereset periodically. As described in the Robust FDI FilterArchitecture section, this periodic resetting enhances the abilityof the robust estimator to track perturbations quickly.300025002000Thrust (lb)150010005000-50011 11.5 12 12.5 13 13.5 14 14.5 15Time (s)Figure 4. Robust filter’s F 1 jet thrust estimates for a nominal and perturbed plant.Robust Failure Detection for Reentry Vehicle Attitude Control Systems 55

45004000350030002500Thrust (lb)2000150010005000-5000 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2Time (s)Figure 5. Jet failure detection using a transient, robust filter F 1 .Table 1. Sum of squared error (10 7 lb 2 ) in jet estimate.Unfailed Jet (steady-state)Nominal Plant (M = 7.5)Perturbed Plant (M = 8.8)Unfailed Jet (transient filter)Nominal PlantPerturbed PlantFailed Jet (steady-state)Nominal PlantPerturbed PlantFailed Jet (transient filter)Nominal PlantPerturbed Plant1.928222.9905.016123.7186.409523.6556.839121.7992.33293.84382.86585.03495.52887.32244.41676.7050For all cases shown in Table 1, whether jets fail or not andwhether a steady-state or transient filter is used, the Kalmanfilter estimates are severely degraded in the presence of plantperturbation. By comparison, the robust filter gives similarnominal and robust performance. What the robust filter givesup in nominal performance when compared with the Kalmanfilter it more than recovers in robust performance.Moreover, the transient robust filter outperforms the transientKalman filter, even for the nominal plant. This is because thehigh-bandwidth Gauss-Markov model of Eq. (2) used to representjet thrusts is only approximate, as no linear model canrepresent a step input. Table 1 also shows that for failuretracking, a transient filter is preferable. When a failure doesnot occur, however, then the steady-state filter gives slightlybetter performance because the transient filter is restartedperiodically, increasing estimation error while it converges toa solution.FAILURE DETECTION AND ISOLATIONWe now discuss the performance of the FDI architecture forfailure isolation. The objective here is to distinguish betweenfailures in the RCS and elevon control surfaces using both filtersin the architecture according to the decision logicdescribed in the preceding section. Figure 6 shows the outputresiduals for a Kalman filter designed to detect failures in theailerons. The jet firing pattern is the same as that of the nominal15-s simulation time. No failure occurs, but elevons areput to use at approximately 6 s. Results for the nominalKalman filter on the nominal plant show that the residualsremain unchanged, i.e., the filter recognized the elevon command.However, the same filter, when used with the perturbedplant, shows a distinct shift in its residual, pointing to afailure that did not occur: a false alarm! Model mismatch and56Robust Failure Detection for Reentry Vehicle Attitude Control Systems

1st measurement (nominal)0.010.0050-0.005-0.010 5Time (s)10 152nd measurement (nominal)0.010.0050-0.005-0.010 5 10 15Time (s)1st measurement (perturbed)0.150.10.050-0.05-0.1-0.152nd measurement (perturbed)0.150.10.050-0.05-0.1-0.15-0.2-0.20 5 10 15 0 5 10 15Time (s)Time (s)Figure 6. Output residuals for the Kalman filter in the absence of failures for the nominal plant (above) and perturbed plant (below).an actual elevon failure may therefore be difficult for theKalman filter to distinguish in the presence of model uncertainty.Raising the threshold is not a satisfactory solution tothe problem of avoiding false alarms because it will result inmissed detections.The robust filter residuals, on the other hand, remain essentiallyunchanged for both the nominal and perturbed plants,as shown in Figure 7. This robust filter design is thereforeinsensitive to model uncertainty.Figure 8 shows the results in the presence of failure whenusing the robust filter design. The elevon is failed on (stuck)for 3.5 s during the simulation period. The robust filter showssimilar shifts in its residuals for both the nominal and perturbedplants. Figure 8 also shows that the same low thresholdmay be set for both plants, thus keeping the rate of falsealarm low.Figure 9 shows the jet estimate when an elevon fails on immediatelyafter a jet failure occurs. Between 0.3 and 0.7 s, one offour jets fails to fire. Between 0.9 and 1.2 s, an elevon remainsfailed on at 1-deg deflection. The shift in the residual of therobust filter concerned with the ailerons F 2 allows for aileronfailure isolation. Subsequently, the filter concerned with jetthrust estimation F 1 is modified to compensate for any sensitivityto the elevon failure. This is accomplished by reparameterizingthe filter. The elevon failure has been modeled as anexogenous input, so that the mapping between that inputand the thrust estimate is simply reduced. As can be seen,adequate jet performance is maintained.CONCLUSIONThis article addressed the problem of failure detection andisolation for next-generation RLV’s attitude control systemsusing the Space Shuttle Orbiter as a model on which to designa prototype system. The objectives were to detect a failurerapidly and to determine accurately whether it occurred in thejet RCS or the aerosurfaces. Correct isolation is essential to thesecond objective, namely, obtaining an accurate estimate ofthe thrust provided by the jets at all times. The problem is particularlydifficult during reentry into the atmosphere becauseof both the rapid variations in Mach number and the largeuncertainties in the vehicle aerodynamic properties.Achieving the preceding objectives required filtering algorithmsthat are, on the one hand, robust to model perturbations,and on the other hand, very sensitive to failures in eitherthe jet firings or the aerosurfaces. A robust FDI architecturethat relies on H ∞ filters has been suggested. Implementingthis filter in simulation has shown that the robust algorithmperforms very well over a wide range of model variations,whereas the nominally optimal Kalman filter can producefalse alarms and leave some failures undetected in the presenceof the same variations.Future work would include looking at the RCS jet propulsionmodel in detail to isolate precisely which jet failed.Robust Failure Detection for Reentry Vehicle Attitude Control Systems 57

1st measurement (nominal)1st measurement (perturbed)0.010.0050-0.005-0.010 5Time (s)0.0110 150.0050-0.0052nd measurement (nominal)2nd measurement (perturbed)0.010.0050-0.005-0.010 5Time (s)0.010.005010 15-0.005-0.010 5 10 15 -0.01 0 5 10 15Time (s)Time (s)Figure 7. Output residuals for the robust filter F 2 in the absence of failures for the nominal plant (above) and perturbed plant (below).1st measurement (nominal)1st measurement (perturbed)0.010.0050-0.005-0.010 5Time (s)0.0110 150.0050-0.0052nd measurement (nominal)2nd measurement (perturbed)0.010.0050-0.005-0.010 5Time (s)0.010.005010 15-0.005-0.010 5 10 15 -0.01 0 5Time (s)Time (s)10 15Figure 8. Output residuals for the robust filter F 2 after an elevon failure is detected by the same filter.58Robust Failure Detection for Reentry Vehicle Attitude Control Systems

45004000350030002500Thrust (lb)2000150010005000-5000 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2Time (s)Figure 9. Jet failure detection using filter F 1 in the presence of elevon failure.ACKNOWLEDGMENTThis research is supported by a Draper Laboratory InternalResearch and Development Project. The authors would alsolike to thank the anonymous reviewers for comments that significantlyenhanced the presentation of the work.REFERENCES[1] Jacquemont, C. M., Aircraft Attitude Determination UsingRobust Estimation, MS Thesis, Department of Aeronauticsand Astronautics, Massachusetts Institute of Technology,Cambridge, MA, August 1997.[2] Mangoubi, R. S., Robust Estimation and Failure Detection: AConcise Treatment, Springer-Verlag, London, August 1998.[3] Appleby, B., Robust Estimator Design Using the H ∞ Normand µ Synthesis, PhD Dissertation, Department of Aeronauticsand Astronautics, Massachusetts Institute ofTechnology, Cambridge, MA, February 1990; also DraperReport T-1065, Cambridge, MA.[4] De Souza, C., U. Shaked, and M. Fu, "Robust H ∞ Filteringwith Parametric Uncertainty and Deterministic InputSignals," Proceedings of the IEEE Conference on Decisionand Control, Vol. 2, Institute of Electrical and ElectronicsEngineers, New York, December 1992, pp. 2305-2310.[5] Xie, L., C. De Souza, and M. Fu, "H ∞ Estimation for Discrete-Time Linear Uncertain Systems," International Journal ofRobust and Nonlinear Control, Vol. 1, No. 2, 1992, pp. 111-123.[6] Mangoubi, R., B. Appleby, and G. Verghese, "RobustEstimation for Discrete-Time Linear Systems," Proceedingsof the American Control Conference, Vol. 1,American Conference on Automatic Control, Baltimore,MD, May 1994, pp. 656-661.[7] Mangoubi, R., B. Appleby, and G. Verghese, "StochasticInterpretation of H ∞ and Robust Estimation," Proceedingsof the IEEE Conference on Decision and Control, Vol.3,Institute of Electrical and Electronics Engineers, New York,1995, pp. 2377-2832.[8] Mangoubi, R., B. Appleby, G. Verghese, and W. VanderVelde, "A Robust Failure Detection and IsolationAlgorithm," Proceedings of the IEEE Conference on Decisionand Control, Vol. 4, Institute of Electrical and ElectronicsEngineers, New York, 1994, pp. 3943-3948.[9] A.D. Rosello, A Vehicle Health Monitoring System for theSpace Shuttle Reaction Control System During Reentry, MSThesis, Department of Aeronautics and Astronautics,Massachusetts Institute of Technology, Cambridge, MA,May 1995.Robust Failure Detection for Reentry Vehicle Attitude Control Systems 59

[10] Gertler, J., "Survey of Model-Based Failure Detection andIsolation in Complex Plants," Control System Magazine,1991, pp. 3-11.[11] Emami-Naeini, A., M. Akhter, and S. Rock, "Effect of ModelUncertainty on Failure Detection," IEEE Transactions onAutomatic Control, Vol. 33, No. 12, 1988, pp. 1106-1115.[12] Patton, R.J. and J. Chen, "Techniques in Robust FaultDetection and Isolation (FDI) Systems," Control andDynamic Systems, Vol. 74, Academic International Press,New York, 1996, pp. 171-190.[13] Frank, P. and X. Ding, "Frequency Domain Approach andThreshold Selector for Robust Model-Based FaultDetection and Isolation," Proceedings of the IFAC/IMACSSymposium on Fault Detection, Supervision and Safety forTechnical Processes (SAFEPROCESS '91),Vol. 1, InternationalFederation of Automatic Control/Society for ExperimentalMechanics, Baden, Germany, 1991, pp. 299-306.[14] Lou, X., A. Willsky, and G. Verghese, "Optimally RobustRedundancy Relations for Failure Detection in UncertainSystem," Automatica, Vol. 22, No. 3, 1986, pp. 333-344.[15] Gertler, J. and D. Singer, "A New Structural Framework forParity Equation Based Failure Detection and Isolation,"Automatica, Vol. 26, No. 2, 1990, pp. 381-388.[16] Horak, D., "Failure Detection in Dynamic Systems withModeling Errors," Journal of Guidance, Control, andDynamics, Vol. 11, No. 6, 1988, pp. 508-516.[17] Emami-Naeini, A., M. Akhter, and S. Rock, "RobustDetection, Isolation, and Accommodation for SensorFailures," Proceedings of the American Control Conference,Vol. 2, 1985, Institute of Electrical and ElectronicsEngineers, New York, pp. 1052-1059.[18] Tsui, C., "A General Failure Detection, Isolation, andAccommodation System with Model Uncertainty andMeasurement Noise," IEEE Transactions on AutomaticControl, Vol. 39, No. 11, 1994, pp. 2318-2321.[19] Saif, M. and Y. Guan, "A New Approach to Robust FaultDetection and Identification," IEEE Transactions onAerospace and Electronic Systems, Vol. 29, No. 3, 1993, pp.685-695.[20] Frank, P., "Enhancement of Robustness in Observer-BasedFault Detection," International Journal on Control, Vol. 59,No. 4, 1994, pp. 955-981.[21] Frank, P. and X. Ding, "Frequency Domain Approach toOptimally Robust Residual Generation and Evaluation forModel-Based Fault Diagnosis," Automatica, Vol. 30, No. 5,1994, pp. 789-804.[22] Mangoubi, R., B. Appleby, and J. Farrell, "Robust Estimationin Failure Detection," Proceedings of the IEEE Conferenceon Decision and Control, Vol. 2, Institute of Electrical andElectronics Engineers, New York, 1992, pp. 2317-2322.[23] Zacharias, G. L., A Digital Autopilot for the Space ShuttleVehicle, MS Thesis, Department of Aeronautics andAstronautics, Massachusetts Institute of Technology,Cambridge, MA, February 1974.[24] De Souza, C., "H ∞ Filtering," Control and Dynamic Systems,Vol. 65, Academic International Press, New York, 1994, pp.323-377.[25] Agustin, R. M., Robust Estimation and Failure Detection forReentry Vehicle Attitude Control Systems, MS Thesis,Department of Mechanical Engineering, MassachusettsInstitute of Technology, Cambridge, MA, June 1998.60Robust Failure Detection for Reentry Vehicle Attitude Control Systems

amses AgustinbiographiesRamses Agustin received a BS in Mechanical Engineering from the University of California, Davis,in 1995. He was awarded a Draper Fellowship, and his research into robust estimation and itsapplication to Space Shuttle reentry formed the basis of his Master’s Thesis. He received an MSin Mechanical Engineering from MIT in 1998, and is currently enrolled in the Graduate School ofBioengineering at the University of California, San Diego.ami MangoubibiographiesRami S. Mangoubi is a member of the Autonomous Control Group at Draper.Since joining Draper in 1983, he has been the technical lead of several IR&D andexternally funded projects in areas such as health monitoring, failure detection,reinforcement learning with application to target tracking, and queuing theoreticalanalysis of computer performance. He has also worked in the areas of optimalcontrol of transport aircraft during cruise flight, simulated annealing for planning methods,image processing, and building protection. He is now developing target geolocation algorithmsfor autonomous vehicles, and flexible mode identification tools for the International SpaceStation. Current activities also include feature-based navigation and high bandwidth adaptivecontrol methods, both for autonomous vehicles, as well as the development of active combustioncontrol methods for liquid fuel space engines. Dr. Mangoubi also developed theory andalgorithms for robust estimation, and introduced the use of these estimators as an approach tothe problem of failure detection in dynamic systems. He supervises MIT graduate students whoare conducting their research at Draper, and is the author of the book Robust Estimation andFailure Detection: A Concise Treatment, Springer Verlag, 1998. He is an invited plenary speaker forthe Year 2000 International Federation of Automatic Control's (IFAC) SAFEPROCESS conference inBudapest, Hungary. He is also currently a member of the IFAC SAFEPROCESS TechnicalCommittee. A senior member of the AIAA, Dr. Mangoubi received an SB in MechanicalEngineering, an SM in Operations Research, and a PhD from the Department of Aeronautics andAstronautics, all from MIT.Robust Failure Detection for Reentry Vehicle Attitude Control Systems: Biographies 61

oger M. HainbiographiesRoger M. Hain received a BS in Aerospace Engineering from Princeton University in 1986. Afterbecoming a Draper Fellow and earning his MS in Aeronautics and Astronautics from MIT in 1988,he worked on numerous space vehicle control and estimation projects at Draper. In addition torobust estimation applications to Space Shuttle Orbiter reentry, these included investigatingvehicle health management techniques for aerospace vehicles, analyzing Orbiter digital autopilotand remote manipulator system dynamic interaction issues for Space Station assemblyflights, and investigating compensation techniques for low-cost, small satellite gyroscopes. He iscurrently employed at the Smithsonian Astrophysical Observatory, where he has participated inthe enhancement and deployment of the Chandra X-Ray telescope's ground attitude determinationand image reconstruction (Aspect Determination) system.eil J. AdamsbiographiesNeil J. Adams is a Principal Systems Engineer in the Systems Engineering andEvaluation Directorate and is the Technology Coordinator for autonomous systemswork at Draper. He is also the Technical Director of the Micro Air Vehicle projectfor the Defense Advanced Research Projects Agency Tactical TechnologyOffice. The Laboratory’s responsibility is to develop the miniature flight controlsystem electronics and the onboard and off-board intelligent planning and control. Mr. Adamscoordinates autonomous systems development for the Small Reconnaissance VehicleDemonstration project for the Office of Special Technology, the Wide Area Surveillance Projectilefor the DARPA Sensor Technology Office, and the Tactical Mobile Robotics project for the DARPATactical Technology Office. Mr. Adams also served as leader of the Control and DynamicalSystems Division from 1995 through 1998. He received a BS from Pennsylvania State Universityand an MS from the University of Cincinnati, both in Aerospace Engineering.62Robust Failure Detection for Reentry Vehicle Attitude Control Systems: Biographies

Jamie M. AndersonPeter A. KerrebrockThe Vorticity Control Unmanned UnderseaVehicle (VCUUV): An Autonomous Robot TunaBased on the paper presented at the 11th International Symposium on Unmanned Untethered Submersible Technology, Durham, NH, August 23-25, 1999Draper Laboratory’s Vorticity Control Unmanned Undersea Vehicle(VCUUV) is the first mission-scale, autonomous underwater vehiclethat uses vorticity control propulsion and maneuvering. TheVCUUV is a self-contained, free-swimming research vehicle thatmimics the morphology and swimming motion of a yellowfintuna. A rigid pressure hull comprises the forward half of the vehicle,which houses batteries, electronics, ballast, and a hydraulicpower unit. The aft section is a freely-flooded articulated robottail terminating in a lunate caudal fin. Using tail kinematic datafrom the Massachusetts Institute of Technology (MIT) RoboTuna,the VCUUV has demonstrated stable, steady swimming up to 2.4kn and aggressive maneuvering trajectories with turning rates upto 75 deg/s. This paper summarizes the vehicle system, fieldexperiments, and performance results of this novel vehiclepropulsion study.ntroductionUnmanned Undersea Vehicle (UUV) technologies haveevolved to produce highly functional and capable platforms fora wide variety of undersea missions. As the supporting technologieshave progressed, so have the mission requirements.Today’s UUV missions require a variety of capabilities, which, insome cases, can be mutually exclusive: high transit speed, longrange and duration, maneuverability, and stationkeeping ability.Fish and marine mammals have captured the interest of vehicledesigners because they can cruise great distances at significantspeed, maneuver in tight spaces, and accelerate and deceleratequickly from rest or low speed with the same integrated propulsionand steering system.In recent years, research in the propulsion and maneuvering flowmechanisms used by fish and marine mammals has demonstratedthe utility of biopropulsion for undersea vehicles. Despiteadvances in UUV technology, little progress has been made inimproving propulsive efficiency and maneuverability. ConventionalUUVs employ a long slender hull with a propeller as themain propulsor and lifting surfaces that provide maneuveringcontrol. Although several recent advanced demonstrations havebeen made with conventional designs, these types of vehiclesare fundamentally limited in their maneuvering performance.Typically requiring several body lengths to execute a turn, thesevehicles can have fatally poor performance at very low speeds.Attempts to improve low-speed performance by using cross-axisthrusters have been effective, but the net result is generally lossof useful hull volume and degraded performance at higherspeeds.This paper presents the results of the first engineering demonstrationsof Draper’s prototype, flexible-hull UUV that propelsand maneuvers like a tuna. Named after the vorticity controlflow control mechanisms employed by fish to propel andmaneuver, the VCUUV mimics the form and kinematics of a largeyellowfin tuna. Across the broad spectrum of fish form andmovement, tuna are most desirable as a vehicle platform; theyare very streamlined, relatively rigid in the forebody, and propelThe Vorticity Control Unmanned Undersea Vehicle (VCUUV): An Autonomous Robot Tuna 63

with low-amplitude movements in conjunction with a highperformancehydrofoil (caudal fin). Vorticity control propulsionis well suited to submarine-based UUVs because itenables similar or greater transit distances and speeds as comparedwith conventional vehicles, and order of magnitudeimprovements in maneuverability. The enhanced maneuveringperformance is particularly desirable during launch andrecovery and in missions that require operation in perturbedenvironments or in close proximity to the free surface andunderwater objects.A common misconception of the biopropulsion concept isthat improved efficiency and maneuvering capability are notworthwhile when compared to the required displacement forthe propulsion system. Attempts to achieve both long-rangeand highly maneuverable rigid-body UUVs have been disappointing,largely due to the need for both propulsion andmaneuvering systems, which rarely operate together effectively.Cross-axis thrusters, for instance, are heavy and detractfrom useful vehicle displacement, and can only be used atzero speed without concern for control nonlinearities (thatcan even include control reversal). For high-speed maneuvers,thrusters are useless; a rigid-body UUV must also have a conventionalpropulsor and control surfaces that cannot produceturning diameters less than several vehicle lengths.In comparison, biopropulsion can provide both efficientpropulsion and maneuvering over a wide range of speedswith a modest investment in vehicle displacement. For example,the VCUUV propulsion system occupies only 23% of thetotal displacement (32% of the envelope displacement due tosome free-flood volume). Thunniform (tuna) morphology andkinematics allow this modest propulsion system displacementwith excellent steady swimming and maneuvering performance.The thunniform model has several advantages for use on anunderwater vehicle system. As with other carangiform fish,the propulsive movements are localized to the last 30 to 40%of the vehicle length and are moderate in amplitude. Tunacaudal fin peak-to-peak excursions rarely exceed 15 to 20% ofthe body length. The localized tail motion allows the forwardbody to be used as a rigid housing for energy, intelligence,payload, etc., without complex flexible pressure hulls or thepower required to move them. The tuna’s fusiform forwardbody has nearly elliptical sections within which internalarrangements are made more easily than with cylindrical sections.The body is low drag, but contains significant packagingspace.The utility of vorticity control propulsion and maneuvering isapparent when one considers the current UUV missions ofinterest. Today’s missions are challenging in that they oftenrequire long transit, long duration on site, loitering withoutloss of power, movement in close proximity to objects fordocking, tagging, etc., and operation in dynamic environmentssuch as shallow waters near the beach zone. Vorticitycontrol propulsion and maneuvering may prove essential inrealizing these missions; fish-like maneuverability may not beas energetically taxing as conventional means of generatinglarge side forces (such as thrusters) in varying conditions. Forexample, fish can loiter at zero speed and rapidly accelerate innearly any direction. Vorticity control propulsion and maneuveringprovide the range and speed of conventional low-draghulls driven by propellers with the added capability of shipdeployedremotely operated vehicles that can maneuver preciselyin a location of interest.BACKGROUNDIn recent years, researchers at MIT have investigated usingrigid flapping foils (combined translation with simultaneousheave and pitch) to generate large propulsion forces at veryhigh efficiencies. Anderson et al. [1] have demonstrated propulsiveefficiencies in excess of 85% with the proper selection ofthe Strouhal number, angle of attack, heave amplitude ratio,and phasing the combined motions. Flow visualization studieswith live fish have shown that their propulsion and maneuveringperformance is related to their ability to control theirwake vorticity. Contrary to bluff bodies towed through thewater, fish generate an oscillatory wake consisting of alternatingvortices arranged in a jet pattern. Manipulating this wakevorticity appears to be a dominant factor affecting the propulsiveperformance of fish. [2]The MIT research culminated in the development ofRoboTuna, a 1.2-m biologically-inspired tow tank model builtto study propulsive efficiency and how it relates to bodymovement. [3],[4] RoboTuna has been exercised in the MITOcean Engineering Testing Tank by prescribing a set of kinematicparameters (angular deflections of each joint, towspeed, phase relationships) and measuring the net powertransmitted to the linkages and the reaction force betweenthe tuna and carriage. Optimal motion was defined as that setof motion parameters that produced no net force on the carriage(self-propelled status) with minimum energy input for agiven speed.Hundreds of combinations of swimming parameters havebeen studied with promising results. The RoboTuna projecthas demonstrated that significant drag reduction can beachieved even without careful tuning of the drive mechanism.The RoboTuna system performed with thrust power ratios inexcess of one and apparent drag reduction up to 70% comparedwith "dead fish" drag. [5] For certain swimming parametersoutside the optimal range, drag amplification up to 300%was observed, which may be used as a very effective brake.VCUUV SYSTEM DESCRIPTIONThe next generation of flexible-hull robots has been developedand tested at Draper (Figure 1). Constructed as a proofof-conceptdemonstration of vorticity control propulsion andmaneuvering, the VCUUV serves as a mission-scale exercise indesign and packaging, as well as a research platform withwhich swimming energetics and maneuvering performanceare evaluated.Although the VCUUV prototype was not intended as anocean-going vehicle, it does contain all the components64The Vorticity Control Unmanned Undersea Vehicle (VCUUV): An Autonomous Robot Tuna

Figure 1. The Draper Laboratory VCUUV.required to execute fully functional autonomous missions:onboard energy, actuation, and control. Vehicle design andfabrication issues addressed in the VCUUV project, includingsizing of actuators, articulated body design, pressure hulldesign, and onboard intelligence, sensors, and power aredescribed in detail in Refs. [6] and [7]. The VCUUV wasdesigned to operate in swimming pool and test tank environmentsat depths less than 10 m at typical live animal cruisespeeds (1 body length per second (L/s)).The need to demonstrate a fieldable vehicle that can serve asa research tool influenced both mechanical and electrical systemdesigns. The mechanical system was required to allowindependent control of tail element oscillation (phase andamplitude of each link) at variable frequencies. The articulatedtail portion of the vehicle required nearly neutral buoyancyso that undesirable rolling moments were not introducedduring oscillation. The hydraulic actuation system wasdesigned to propel the vehicle at 2.5 kn for up to 3 h of testing.The tail articulation range was based on the requirementfor a desired turning circle diameter of 2.5 lengths or less.Although designed for modest depths (

Figure 2. VCUUV system layout.sensors (internal temperature, hydraulic pressure, leak detection,etc.). All sensors, except those dedicated to the tail linkage,are located in the electronics assembly inside thepressure hull. In addition, an external speed sensor was usedin some experiments. By measuring the linkage force and displacementas functions of time, the instantaneous powerabsorbed by the tail assembly can be calculated as the productof force times velocity for each link.The requirements for control of the vehicle were precise, androbust control of the tail linkage was such that its movementsadhered to the prescribed linkage trajectory. Closed-loopcontrol of the vehicle position was not implemented as thegoal of the experiments was to identify open-loop characteristicsthat could be used later to design a closed-loop trajectorycontroller. Heading control was explored throughcompass feedback on a variety of trajectories. Depth controlwas tested but not implemented as the vehicle was passivelystable in depth due to careful ballasting.The key parameters in thunniform steady swimming are vehiclespeed, oscillation frequency and amplitude, propulsivewavelength, shape parameters that describe the envelope oftransverse motion, tail phasing, and angle of attack. [3] Forsteady, straight swimming, a traveling propulsive wave isrequired that moves from head to tail with increasing amplitudenear the tail. The kinematics are described in detail inRefs. [3], [6], [7], and [8]. Turning kinematics are less understood,and consequently, the VCUUV has most of its designmargin in the turning requirements (amplitudes and loadingconditions).As in actual tuna, dive plane control is provided by two pectoralfins located near the midbody. Real tuna, which are negativelybuoyant, use the pectoral fins for speed-dependent liftto control vertical position in the water column. At low speeds,the fins are fully splayed for maximum lift; as the speedincreases, the fin sweepback angle is increased until the finsare fully adducted, effectively reducing the lifting area. [8]The VCUUV pectoral fins are positioned in the same locationas real tuna pectoral fins, slightly forward of the center of gravity.Thus, the pectoral fins act as canards. Unlike actual tuna,which can articulate their pectoral fins in pitch and sweepangle, the VCUUV pectoral fins only pitch, controlled by torquemotors located inside the pressure hull. For simplicity, thepectoral fin shapes are swept and tapered NACA 0015 sectionsroughly matching the chord and span of actual tuna fins.Because of the large metacentric height of the vehicle, pectoralfin lift causes little change in the vehicle while controllingdepth.FIELD TRIALSThe VCUUV has completed 20 days of testing and approximately100 autonomous experiments, beginning with thevehicle’s first swim in April 1998 in a local swimming pool.Preliminary trials were completed at the University of NewHampshire (UNH) Ocean Engineering Laboratory in a 60-ft x40-ft x 20-ft-deep freshwater tank. Autonomous swimmingand simple maneuvers were achieved early in the experimentswith no physical or electrical connections to theground station.Early field trials (April-June 1998, at the UNH test tank) focusedon engineering diagnostics and adjustments, especially in tuningthe hydraulic system and control system performance.Hydraulic pressure, control gains, and the desired tail kinematics66The Vorticity Control Unmanned Undersea Vehicle (VCUUV): An Autonomous Robot Tuna

were adjusted to give good tracking performance for frequenciesup to 1 Hz. Tail oscillations above 1 Hz were not pursueddue to larger-than-expected forces that led to poorcontrol performance. Simple maneuvers were explored,including coasting turns and biased swimming movementresulting in circular trajectories. Because of the short distancesavailable in the UNH tank, steady-state swimming atterminal velocity could not be achieved. Thus, the field trialswere moved to local freshwater lakes for longer-duration andhigher-speed tests.In the Summer and Fall of 1998, the VCUUV was tested in twolocal freshwater lakes (Hopkinton State Park, Hopkinton, MAand Nickerson State Park, Brewster, MA). These sites were chosenfor their proximity to Draper and their clean, undisturbedfresh water. The lake experiments took advantage of the largerbodies of water by swimming longer and more aggressivemissions. A systematic study of tail kinematic parameter variationswas completed, as well as several basic open- andclosed-loop horizontal plane maneuvers. Experiments inheading control were conducted yielding straight, circular,square, and zigzag trajectories. Swimming kinematic parameterswere varied for straight swimming and maneuvering toidentify the key parameters and how they affect swimmingperformance.A typical experimental run began with the vehicle at rest andsubmerged while baseline inertial data were collected. Thevehicle then began the swimming tail motion that acceleratedthe vehicle to steady-state velocity. The typical accelerationperiod was 20 to 30 s for a 1-Hz tail oscillation. Afterswimming a total of 1 min, the vehicle coasted to a stop whilethe tail was held in a straight position. The tail movementclosely adhered to the kinematics reported for optimal performance.[3]STRAIGHT SWIMMING RESULTSThe maximum speed attained in the field trials was 1.25 m/s(2.4 kn) at a 1-Hz tail oscillation while under heading control(closed-loop control using compass feedback). Without headingcontrol, the maximum speed was 1.19 m/s (2.3 kn). Figure3 shows the steady-state speed as a function of frequency, aswell as the speed predicted by extrapolating the RoboTunaresults to the VCUUV vehicle size. The published RoboTunaresult is at a single speed, which is also marked in Figure 3. TheVCUUV speed was measured by both speed sensor and timetrials with very good agreement between the two measurements.The speed increases with frequency up to 1 Hz, at which pointthe tail kinematic performance suffers and the speed decreasesdue to the hardware limitations of the hydraulic system.The actual vehicle speed was less than that predicted by 16%,which may be due to differences in the kinematics betweenthe VCUUV tail motion and that of the RoboTuna. Using thesame length reference (pre-peduncular length), the VCUUVachieved 0.61 L/s, whereas the RoboTuna achieved 0.65 L/s.Unfortunately, degraded performance above 1 Hz did notallow the vehicle to achieve the design speed of 1.0 L/s.21.5U (m/s)10.500 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6f (Hz)Figure 3. VCUUV steady-state speed as a function of tail oscillation frequency.The Vorticity Control Unmanned Undersea Vehicle (VCUUV): An Autonomous Robot Tuna 67

MANEUVERING RESULTSThe VCUUV possesses excellent maneuverability that matchesthe performance reported for live animals. [9] Several maneuveringtrajectories were explored, including abrupt discreteturns and continuous turning, which yielded circular and spiraltrajectories. Small stationary turning circles (8- to 10-ftdiameter) were achieved easily by simple biasing of thestraight swimming tail motion.Figure 4 illustrates the heading angle measured during acoasting turn with full body deflection executed at full speed(1.2 m/s). Two measurements are indicated in the figure: theintegral of the heading gyro yaw rate and the absolute headingangle measured by the compass. The data from both sensorsare in agreement, and indicate that the vehicle turnedthrough approximately 180 deg in less than 10 s. The maximummeasured heading rate during this maneuver was 75deg/s. By comparison, a conventional UUV with control surfacesturns at approximately 3 to 5 deg/s, often requiring severalbody lengths and as long as 1 min to complete a 180-degturn. This radical improvement in maneuverability mayenable mission elements previously considered beyond thecapabilities of UUVs, including close inspection and rapidcourse change tasks, operation in highly perturbed environments,and launch and recovery to a moving submarine.Figure 5 illustrates the heading angle achieved in an aggressivezigzag trajectory. The vehicle swam straight for 20 s toattain steady-state velocity and then swam zigzag legs at ±45deg from the initial heading with 10 s spent on each leg. Thevehicle maintained forward velocity during the maneuver at2.4 kn. The maximum yaw rate for this maneuver was approximately30 deg/s, which also outperforms conventional systemsby several factors. The zigzag maneuver indicates thelevel of performance that can be achieved in simple searchpatterns or in close inspection missions. The vehicle can makeaggressive course changes in-stride while preserving forwardspeed.CONCLUSIONDraper’s VCUUV is the first autonomously operated fish-likeUUV that is sized realistically for real-world missions and payloadand able to move with arbitrary thunniform kinematics.The fish propulsion paradigm offers the order-of-magnitudeimprovement in maneuvering capability required for today’schallenging UUV missions in dynamic, cluttered environments.The VCUUV prototype proves the concept while servingas a research testbed. Lessons learned from the VCUUVcan be applied to new vehicles mechanically optimized toachieve fish-like capabilities in engineered vehicles.Further work will demonstrate that this novel propulsionmechanism can be incorporated into vehicles capable of practicalmissions that are sufficiently compelling to justify the riskof their development. This effort will include demonstratingthe scalability of vorticity-control propulsion (both up anddown) and developing more stealthy actuators, three-axismaneuvering capability, and sensors and control systems tosupport fully closed-loop propulsion and maneuvering control.200150100Heading (deg)500-50-1000 20 40 60 80 100 120 140 160Time (s)Figure 4. Heading during a hard turn.68The Vorticity Control Unmanned Undersea Vehicle (VCUUV): An Autonomous Robot Tuna

80604020Heading (deg)0-20-40-60-800 50 100 150 200 250Time (s)Figure 5. Heading during zigzag maneuvers.ACKNOWLEDGMENTSThis project has been funded internally by The Charles StarkDraper Laboratory, Inc.REFERENCES[1] Anderson, J.M,. K. Streitlien, D.S. Barrett, and M.S.Triantafyllou, "Oscillating Foils of High PropulsiveEfficiency," Journal of Fluid Mechanics, Vol. 360, 1998, pp.41-72.[2] Anderson, J.M., Vorticity Control for Efficient Propulsion,PhD Thesis, Massachusetts Institute of Technology/Woods Hole Oceanographic Institution Joint Program,1996.[3] Barrett, D.S., Propulsive Efficiency of a Flexible HullUnderwater Vehicle, PhD Thesis, Massachusetts Institute ofTechnology, 1996.[4] Triantafyllou, M.S. and G.S. Triantafyllou, "An EfficientSwimming Machine," Scientific American, March 1995.[5] Barrett, D.S., M.S. Triantafyllou, D.K.P. Yue, M.A.Grosenbaugh, and M.J. Wolfgang, "Drag Reduction inFish-Like Locomotion," Journal of Fluid Mechanics, 1999,accepted.[6] Anderson, J.M. and P.A. Kerrebrock, "The Vorticity ControlUnmanned Undersea Vehicle – An Autonomous VehicleEmploying Fish Swimming Propulsion and Maneuvering,"Proceedings of the 10 th International Symposium onUnmanned Untethered Submersible Technology, Durham,NH, September 7-10, 1997, pp. 189-195.[7] Anderson, J.M., P.A. Kerrebrock, and M.S. Triantafyllou,"Concept Design of a Flexible-Hull Unmanned UnderseaVehicle," Proceedings of the 7 th International Offshore andPolar Engineering Conference,Vol. II, Honolulu, Hawaii, May25-30, 1997, pp. 82-88.[8] Dewar, H. and J.B. Graham, "Studies of Tropical TunaSwimming Performance in a Large Water Tunnel, Part III:Kinematics," Journal of Experimental Biology, Vol. 192,1994, pp. 45-59.[9] Blake, R.W., L.M. Chatters, and P. Domenici, "Turning Radiusof Yellowfin Tuna (Thunnus albacares) in UnsteadySwimming Manoeuvres," Journal of Fish Biology, Vol. 46,1995, pp. 536-538.The Vorticity Control Unmanned Undersea Vehicle (VCUUV): An Autonomous Robot Tuna 69

amie M. Andersonbiographies biographiesJamie M. Anderson is a Senior Member of the Technical Staff and has been withDraper Laboratory since 1996. She is the Group Leader for Vehicle Systems in theMechanical and Instruments Division. She is Principal Investigator for the VCUUV,a prototype fish-like autonomous underwater vehicle designed to study thepropulsion and maneuvering characteristics of flexible hull vehicles. She managedand supervised all aspects of vehicle concept proposal, design, fabrication, budget, andstaff. Dr. Anderson provided reports to management, presented conference papers, and interfacedwith popular press and educational programs. She is also Mechanical Design Task Leaderfor the Wide Area Surveillance Projectile (WASP) project, an autonomous aerial vehicle that isdeployed by launching inside a munition subjected to 15,000 times the acceleration of gravity.Previously, Dr. Anderson was a research assistant in the Ocean Engineering Department at MIT.Dr. Anderson has co-authored several papers for the Journal of Fluid Mechanics and conferencepublications. She received a BS in Mechanical Engineering in March 1989 from the University ofCalifornia, San Diego and MS and PhD degrees in Oceanographic Engineering from MIT andWoods Hole Oceanographic Institution in February 1992 and February 1996, respectively.eter A. KerrebrockPeter A. Kerrebrock is a Member of the Technical Staff in the Vehicle SystemsGroup. He is responsible for performing mechanical design and analysis tasks ina variety of subjects related to undersea, ground, and aerial systems. Mr.Kerrebrock recently served as Lead Mechanical Engineer and Naval Architect onthe VCUUV Project, and is currently Mechanical Task Leader for the Tactical MobileRobotics and Autonomous Systems IR&D projects. Prior to that, he served as an Alternate LeadMechanical Engineer in the DARPA AMMT UUV program. He also recently co-led Draper’smechanical and naval architectural efforts in the LMRS IPT and COEA, and the LMRS UBD designeffort that followed. Previously, Mr. Kerrebrock was a contributor to the SOMSS and AVC programsfor the development of submarine torpedo tube launchable UUVs. He also served as PrincipalInvestigator of an IR&D project to investigate advanced energy storage and conversion systemsfor UUVs, and as Mechanical Task Leader for Draper’s concerns in the U.S. Navy DSRV program.Before joining Draper Laboratory in 1989, Mr. Kerrebrock was a Mechanical Engineer at the U.S.Naval Undersea Warfare Center (NUWC) in Newport, Rhode Island, where he was involved in thedevelopment of thermal torpedo propulsion systems. During the torpedo MK 48 ADCAP program,Mr. Kerrebrock directed all propulsion system dynamometer testing, evaluation, and failureanalysis from pre-ADM through FSED. Mr. Kerrebrock received a Sustained Superior AchievementAward for his work on the ADCAP program.70The Vorticity Control Unmanned Undersea Vehicle (VCUUV): An Autonomous Robot Tuna: Biographies-- ---- -

Marc S. WeinbergII:"'JI.m-4..... 7.~QI--4.A/ -l.m-41T__ 7.:~[ '"Working Equations for PiezoelectricActuators and Sensors©1999 IEEE. Reprinted, with permission, from the ASME/IEEE Journal of MEMS, Vol. 8, No. 4A new solution to the force, displacements, and charges developedin piezoelectric beams is derived. Differing from previous solutions,this development determines the neutral axis where thebending strains are zero and results are in a closed-form solution(without matrix inversion). With the closed form, simplificationsbecome evident that increase understanding and facilitate calculations.These equations are then expanded to account for axial,built-in strains in the beam. A design example where axial forcesexerted by the piezoelectric layer are important is presented.z axisi’th layerRM L LF Lz iz Lθ Lx axisNomenclature for multimorph analysis with one free and one builtinend.PntroductionSmits and Choi [1] derive the dynamic and static constituentequations for a structure that contains two layers, one of a piezoelectricand a second of an elastic material. DeVoe and Pisano [2]extended the analysis to a multimorph, a cantilever that consistsof several layers of dielectric, piezoelectric, elastic, and conductors.These analyses trace back to Timoschenko’s bimetallicbeam. [3]This paper presents an alternate development that adds furtherinsight into piezoelectric design by calculating the neutral axis, [4]where the bending strains are zero, and results in a closed-formsolution (without inverting matrices with an order that is thenumber of layers plus one as in Ref. [2]). With the closed form,simplifications become evident that increase understanding andfacilitate calculations and rough estimates. These equations arethen expanded to account for axial, built-in strains in the beam.Draper Laboratory has used these relations since 1985 to designquartz vibrating gyroscopes and accelerometers, and silicon andpiezoelectric hydrophones, microphones, and flexural plate wavesensors. [5]-[8]Section 1 derives the curvature and deflections in a multilayerbeam with no axial strains. In Section 2, the charge caused bybeam deflection is derived. This is used as a check in Section 3,where the lumped parameter equations for a piezoelectric beamare developed. The results are extended to include axial tensionor compression in Section 4. Section 5 presents a design examplewhere axial forces exerted by the piezoelectric layer areimportant. Conclusions are listed in Section 6.•- ---- -oWorking Equations for Piezoelectric Actuators and Sensors 71

1. MULTILAYER BEAM MODEL WITH PIEZOELECTRIC ANDTHERMAL EFFECTSASSUMPTIONSFollowing Ref. [2], the geometry of an m-layer multimorph issketched in Figure 1. As in Ref. [2], the following is assumed:(1) The layers can be either piezoelectric or purely elastic.(2) The interfaces between layers are continuous and do notslip with respect to one another.(3) Each element of the beam is in static equilibrium.(4) Beam thickness is much less than the radius of curvatureinduced by torque, piezoelectric, or thermal effects.(5) The beam width is comparable to thickness so that allstresses are in the xz plane.(6) If the beam is much wider than the thickness, the beamcan be considered to be in-plane strain with e y = 0. If thebeam is isotropic, the Young modulus and piezoelectriccoefficients used below can be replaced by:E i å E i / (1 – ν 2 i) (1)d 31i å d 31i (1 + ν i ) (2)whereν = Poisson’s ratiod 31 = piezoelectric coupling coefficienti= index for a material layer(7) Piezoelectric coupling is small. The coupling is a dimensionlessquantity defined as k 2 = d31E 2 / ε where ε is thedielectric constant. When k 2 is small compared to 1, theelectrical capacitance and mechanical stiffness can be calculatedwithout considering the piezoelectric effect. Thecoupling coefficient k 2 is roughly 1% for zinc oxide andaluminum nitride and 12% for lead zerconium titanate(PZT).(8) The y and z axes defined in Figure 1 are principal axes ofthe area cross sections.(9) The initial treatment, Sections 1 through 3, considers smallaxial stresses. Residual stresses are treated in Section 4.The thin beam and continuity between layers allow the axialstrain in any layer to be defined as:z axisi’th layerLFigure 1.RM L F Lθ Lz Lz ix axisNomenclature for multimorph analysis with one free and onebuilt-in end.In Refs. [1] through [3], Eq. (3) is implicit in the strains definedat the interfaces between layers. Solutions for z N and R arenow derived. From the stress-strain relationships, the axialstress is:σ i = E i (e x – d 31i E i – α i ∆T) (4)whereE = Young’s modulusE = z component of electric field (agreeing with Ref. [2],electric field is bold)d 31 = piezoelectric coefficient coupling z electric field to xstrainα = thermal expansion coefficient∆T = temperature variation from nominali = index denoting material layerWith a positive electric field E and coefficient d 31 , an unrestrainedbeam will lengthen. If the piezoelectric material is inthe upper part of the beam, the unrestrained beam will curvedownward.GENERAL FORM OF RADIUS OF CURVATURETemperature and electric field have similar roles so that thethermal term can be omitted without losing generality.An externally applied axial force P must be the integral of theaxial stress taken over the cross section at any axial position.Using Eqs. (3) and (4):Pwherez Ne x= neutral axis, defined from an arbitrary reference= axial strain(3)where A I = cross section area of layer i.Perform the integration indicated in Eq. (5):(5)R= radius of curvature, often a function of axial position;the sign convention is defined in Figure 1(6)72Working Equations for Piezoelectric Actuators and Sensors

where z i is measured from the arbitrary reference to the centerof area of area i. All quantities except z N and R are known.At any axial position, the external moments must be balancedby the internal stresses.If the reference is selected as the composite beam’s neutralaxis,Σ iz i E i A i = 0 and the curvature for torque inputs Eq. (12)becomes the classic result: [4] (13)(7)where I = the area moment of inertia about each layer’s centerof area.Solve Eqs. (6) and (8) simultaneously to obtain the neutral axisand the radius of curvature in closed form.(8)(9)(10)When M and P equal zero, the numerical results for curvatureare identical to those obtained by DeVoe’s matrix inversion. [2]When only one elastic and one piezoelectric layer are used,the curvature identically matches DeVoe [2] and Smits. [1]WORKING EQUATIONSUnlike Ref. [2], Eqs. (9) and (10) do not require inversion of anorder m+1 square matrix and matrix multiplication, and areoften more amenable to available mathematical tools.Moreover, the solutions for radius of curvature and neutralaxis can be reshaped to be much more useful in engineeringpractice and lead to piezoelectric equations with axial stresses,another contribution of this paper.From Eq. (9), the curvature is the superposition of torque, force,and electric field effects. Consider the neutral axis for torqueinputs only; that is, no axial force or z electric fields.(11)where z M denotes the neutral axis for torque inputs, a conceptwell known in conventional stress analysis. [4] This neutral axisis the weighted center of EA for the entire composite beamand does not depend on the magnitude of the torque. FromEq. (9), the curvature per unit torque C Mis:(12)whereI = the area moment of inertia for each layerZ i = z i - z M = vertical position measured to the torqueneutral axisConsider axial force P. Since the distances z are measured froman arbitrary reference, a torque M = -P(z M + ∆z) is also applied.The distance ∆z is between the torque neutral axis and theapplication of the force. With zero electric field, the radius ofcurvature per force inputs C P is determined from Eq. (9).C P = -C M ∆z (14)For obtaining curvature, the axial force can be treated as amoment -P∆z, an important result used to extend results tosituations with axial tension in Section 4.The neutral axis for force inputs is determined with zero electricfield and M = -P(z M + ∆z):(15)Determine the axial strain along the torque neutral axis by Eq.(3) and use Eqs. (11), (12), (14), and (15):(16)The strain determined by Eq. (16) is that obtained by consideringparallel beams restrained to only x displacement. Thestrain is the sum of the axial term Eq. (16) plus a term dependingon the torque -P∆z operating on the curvature C M .Next consider the curvature from electric fields. Let M = 0 andP = 0. Let the reference axis be the torque neutral axis so thatΣ iz i E i A i = 0. From Eqs. (9) and (13), the curvature caused bythe z direction electric fields is:(17)where Z i is measured with respect to the torque neutral axis.Each piezoelectric layer acts as though it exerts a forceE i A i d 31i E i at a moment arm Z i from the torque neutral axis. Asdescribed after Eq. (4), with a positive electric field E and coefficientd 31 , an unrestrained beam will lengthen. If the piezoelectricmaterial is in the upper part of the beam, theWorking Equations for Piezoelectric Actuators and Sensors 73

unrestrained beam will curve downward in agreement withEq. (17). The axial strain at the torque neutral axis is determineddirectly from Eq. (16).If the conducting plates span just one dielectric or piezoelectriclayer, the electric field is given by:whereV(18)= potential across plates (assume voltage is applied sothat electric field E is positive. Generally E = -DV) [9]t = thickness of dielectric layerWhen the plates span several layers, as does the example inRef. [2], the electric field in each layer is:where j spans the layers between the electrode layers.2. PIEZOELECTRIC SENSING(19)The charge across grounded plates is derived for a torqueinput. The reasons for grounded plates will become apparentin Section 3. For a piezoelectric beam, the companion of Eq.(4) that describes the electric field displacement is:D i = ε i E i + d 31i σ i (20)whereD = electric displacement in z directionε = permittivityσ = x direction normal stressThe subscript i indicates the layer. Consider just one layer andomit the index. Within a rectangular layer, the electric displacementis constant. For a torque input, the stress is givenby Eqs. (3) and (4). Write Eq. (20) to emphasize the electricfield:(21)With no charge buildup, the integral of E must be zero sinceboth plates are grounded; therefore:The charge on the plates is determined by:whereQ = DbL (23)b = beam widthL = beam lengthWith Eqs. (22) and (23), the charge as a function of torque is:Q = d 31 EC M M(z M – z i )bL (24)When the electrodes enclose more than one piezoelectric layer,the charge is given by:(25)where the index j describes the layers between the electrodes.Between electrodes A j = bt j .3. LUMPED PARAMETER MODELS - NO AXIAL TENSIONSimilar to Ref. [1], derive a lumped parameter model where theforce and displacements are described at the tip of the beam.Assume zero displacement and angle at z = 0. Since curvatureC = d 2 z/dx 2 and θ = dz/dx, the curvature can be integrated todetermine the tip deflections. As in Ref. [1], assume that thebeam is uniform along its length L.whereM V = torque per unit voltage across electrodes=(26)wheretz i(22)= layer thickness= distance from reference to piezoelectric’s center ofEAF V = axial force per unit voltage=74Working Equations for Piezoelectric Actuators and Sensors

L = beam lengthC = capacitance of electrodes = εbL/t for a single layerbetween electrodes; Ref. [1] had a similar expressionthat differed by the coupling k 2 (Section 2) andcan be considered identicalThese results for displacement versus torque and force matchthose tabulated in Ref. [10], Table 9. The results for compressionare obtained by setting P to a negative number in Eq. (27).k x = = axial stiffnessBy Maxwell’s reciprocity, the matrix in Eq. (26) should be symmetric.The [2,4] term relating θ L to V was derived in Eqs. (17)and (19). The [4,2] term relating charge to torque was derivedin Eq. (25). The two terms’ being equal is a partial check on thederivations. Equation (25) is compatible with Section 3, wherethe voltage across the sensing electrodes was assumed zero.4. AXIAL TENSION OR COMPRESSIONThe results of Section 2 enable constructing a lumped parametermodel when in-plane stresses are built into the beam.From Eqs. (13), (14), and (17), the vertical deflections with builtinstresses are described by:whereEI =(27)= Young’s modulus time inertia defined with respectto torque neutral axis, Eqs. (11) and (13)In Eq. (27), the torques are defined with respect to the torqueneutral axis Eq. (11). Because of the curling of the beam, theaxial load P at the end of the beam torques the beam eventhough P is applied at the neutral axis. For a uniform beam,the torque M L and the piezoelectric terms have similar effect.Solve Eq. (27) with the boundary conditions z(0) = 0, z’(0) = 0,and z(L) = z L . Inserting x = L into the solution of Eq. (27), thelumped parameter model [analogous to Eq. (26)] for a beamunder axial tension is:where k =for tension.(28)where k =5. APPLICATIONSfor compression.(29)DeVoe compares the model versus measured data for a cantilever.Draper has obtained similar results for PZT depositedon silicon and aluminum nitride deposited on silicon nitride.Because the quality of the deposited piezoelectric film varies,Draper’s results are similar to DeVoe’s in that a fit for the piezoelectriccoefficient d 31 is often needed.Figure 2 is an example that is applicable to oscillators, vibratingaccelerometers, and approximately to audio speakerdiaphragms. Consider a beam with a total length of 4L. Thebeam is built-in at both ends and consists of silicon, PZT, gold,and oxide (see Table 1 for a summary of their characteristics).The gold electrodes cover the outer quarters of the beam.(One electrode covering the entire beam will not causemotion.) The silicon is considered ground. A nitride film coversthe gold.For 5-µm-thick silicon, 3-µm PZT, 1 V applied across the PZT,and L = 125 µm, the axial force is 0.00283 N (50% calculated byEqs. (26), (28), or (29)). The torque neutral axis is 3.2 µm belowthe center of the PZT layer. The torque exerted by the PZT is1.8 × 10 -8 N-m, defined with respect to the torque neutral axis.A quick estimate of available force is this piezoelectric torquedivided by the beam length (125 µm), a calculation that gives144 µN. For zero displacement at the beam center, the maximumforce at the beam’s center (Figure 2) is 220 µN. The maximumdeflection is 0.057 µm with F equal zero. For theseconditions, the force and displacements can be calculated byEq. (26) without considering the axial force. The stiffness F/z is3860 N/m.As dimensions and drive voltage vary, the force and displacementcalculations may warrant including the axial force as inEqs. (28) and (29). Figure 3 plots the stiffness F/z (Figure 2) asthe silicon thickness and beam length are varied. Results withaxial force are divided by the no-axial-force calculations.Conditions are listed in Table 1. Excitation is 10 V and PZTthickness is 3 µm. For 8-µm thickness, models with axial forceapproach the no-axial-force situation. As the beams becomeWorking Equations for Piezoelectric Actuators and Sensors 75

F, zPZT-gold-oxide-~ ~~SiiL/4Figure 2. Multimorph with built-in ends.Table 1. Parameters for multimorph example.Heightµmt1 to 830.10.1Widthµmb500500500500Young’s ModulusN/m 2E1.68 × 10 116.1 × 10 107.5 × 10 107.9 × 10 10Poisson’s– – –ν0.0650.280.420.3Piezo CoefficientC/Nd 31-1.7 × 10 -10Dielectric Constant– – –ε1700thinner or longer, the stiffness with axial tension increases andthe stiffness with compression decreases. When kL is of order0.1, the linear relationships in Eq. (26) lose accuracy. When thecompression curve reaches zero stiffness, buckling occurs.76Working Equations for Piezoelectric Actuators and Sensors6. CONCLUSIONS10'--------The piezoelectric effect and other in-plane forces, such asthermal expansion, can be modeled as additional torques referencedto the torque neutral axis. This result greatly simplifiesprevious models. A curvature per unit external torque and-- --- -

3.02.52.0F/z ratio to linear1.51.00.50.00e+0 2e-6 4e-6 6e-6 8e-6Si thickness (m)Figure 3. Relative stiffness for sample multimorph of Table 1.a piezoelectric torque, which is simply calculated, define thepiezoelectric situation. This formulation extends the piezoelectriclumped parameter model to situations with in-planetension and compression.In a micromechanical multimorph, the axial stress caused bythe piezoelectric can be sufficient to alter stiffness from thatcalculated through linear models.REFERENCES[1] Smits, J. and W.-S. Choi, "The Constituent Equations ofPiezoelectric Heterogeneous Bimorphs," IEEETransactions on Ultrasonics, Ferroelectrics, and FrequencyControl, Vol. 38, No. 3, 1991.[2] DeVoe, D. and A. Pisano, "Modeling and Optimal Design ofPiezoelectric Cantilever Microactuators," Journal ofMicroelectromechanical Systems, Vol. 6, No. 3, 1997.[3] Timoschenko, S., "Analysis of Bi-metal Thermostats,"Journal of the Optical Society of America and Review ofScientific Instruments, Vol. 11, No. 3, 1925, pp. 233-256.[4] Crandall, S. and N. Dahl, An Introduction to the Mechanicsof Solids, McGraw-Hill Book Co., New York, 1959, pp.285-295.[5] Petrovich, A., M. Weinberg, J. Williams, Quartz MonolithicAccelerometer, U.S. Patent 5,315,874, May 31, 1994.[6] Weinberg, M., Quartz Tuning Fork Gyroscope, 5,388,458,U.S. Patent, February 14, 1995.[7] Kourepenis, A., A. Petrovich, M. Weinberg, "Developmentof a Monolithic Quartz Resonator Accelerometer,"Proceedings of the Fourteenth Biennial Guidance TestSymposium, Holloman AFB, NM, October 2-4, 1989.[8] Kourepenis, A., A. Petrovich, M. Weinberg, "Low CostQuartz Resonant Accelerometer for Aircraft InertialNavigation," Transducers '91, Digest of Technical Papers,International Conference on Solid-State Sensors andActuators, June 24, 1991, IEEE Cat. #91CH2817-5.[9] Woodson, H. and J. Melcher, Electromechanical Dynamics,Vol. I, J. Wiley & Sons, Inc., New York, 1968, p. B10.[10] Roark, R.J., W. Young, Formulas for Stress and Strain, FifthEdition, McGraw-Hill Book Co., New York, 1975, Tables 9and 10.Working Equations for Piezoelectric Actuators and Sensors 77

arc S. WeinbergMarc S. Weinberg received BS, MS, and PhD degrees from the MassachusettsInstitute of Technology in Mechanical Engineering. He is Laboratory TechnicalStaff and Group Leader in the Systems Engineering and Evaluation Directorate.He is responsible for the design and testing of a wide range of micromechanicalgyroscopes, accelerometers, hydrophones, microphones, angular displacementsensors, chemical sensors, and biomedical devices. He holds 19 patents with 14 additional inapplication. He was given Draper’s Best Patent, Best Publication, and Distinguished PerformanceAward for his work on the tuning-fork gyro, the first silicon micromechanical gyroscope todemonstrate resolution better than 100 deg/h in 60 Hz.biography78Working Equations for Piezoelectric Actuators and Sensors: Biography

The following pages contain the bibliographical information and a brief abstract of additionalpapers that have been formally published by Draper engineers during the 1999 calendar year.Ash, M.E.; Trainor, C.V.; Elliott, R.D.; Borenstein, J.T.; Kourepenis, A.S.; Ward,P.A.; Weinberg, M.S.Micromechanical Inertial Sensor Development at Draper Laboratory with Recent TestResultsSymposium Gyro Technology 1999, Stuttgart, Germany, September 14-15, 1999, Proceedings(A00-17839 03-35), Stuttgart/Bonn, Universitaet Stuttgart/Deutsche Gesellschaft fuer Ortungund Navigation, 1999, pp. 3.0-3.13A Coriolis Vibratory Gyroscope (CVG) detects inertial rotation bymeans of the Coriolis effect acting on a vibrating mass.The DraperMicroelectromechanical Systems (MEMS) CVG design uses two siliconproof masses that are capacitively comb-driven to vibrate180 deg out of phase suspended over a glass plane substrate towhich the silicon structure is anodically bonded. Angular motionabout an axis perpendicular to the velocity vector causes out-ofplanemotion of the proof masses, which is capacitively read outand amplified to give a measure of the inertial angular velocity.Through a continuing process of design improvements, theDraper CVG bias stability now surpasses 10 deg/h in temperaturecontrolled(±0.25°C) 6-h drift tests. Temperature-compensatedperformance over -40°C to +85°C is 50 deg/h and 200 ppm forbias and scale factor, respectively. Angle random walk of 0.25 deg√h is typical, with best-to-date performance of 0.05 deg √hobserved. A companion MEMS torque-rebalance accelerometerhas also been developed with an unbalanced silicon teeter-totterproof mass suspended over a glass ground plane that is providedwith pickoff and electrostatic torquing capacitors. It has demonstratedsub-milli-g performance in temperature-controlled(±0.25°C) 6-h drift tests.Barton, G.H.; Tragesser, S.G.Autonomous Intact Abort System for the X-34Atmospheric Flight Mechanics, held in Portland, OR. Sponsored by: AIAA, August 9-11,1999Autonomous algorithms are developed that provide trajectoryguidance for horizontally landing vehicles such as the X-34 undera variety of abort conditions. The nominal guidance system of theX-34 is incapable of directing the vehicle to a safe landing formany possible situations in which the trajectory is far away fromnominal conditions (as in the case of an engine failure). To minimizethe risk of losing the vehicle, the autonomous intact abortsystem considers multiple landing sites and redesigns certainguidance inputs in order to adapt to the new conditions presentedby the abort. The abort system design is demonstrated in ahigh-fidelity simulation to prove the feasibility of the concept forvarious engine-out scenarios. These abort algorithms are beingincorporated into the X-34 vehicle to flight test this new technologyas a part of the Future-X Pathfinder Flight Demonstration Program.Bedrossian, N.S.; McCants, E.Space Station Attitude Control During Payload OperationsAstrodynamics Specialist Conference, Anchorage, AKEvaluating the feasibility of planned robotic operations requiresan analysis methodology and tools that can assess proposed attitudecontrol strategies quickly. In this paper, an efficient approachto model the attitude dynamics of the Space Station during payloadmotion is presented. This formulation was then used todevelop momentum optimal attitude command trajectories forthe Space Station Control Moment Gyroscope (CMG) attitudehold controller for use during robotic payload operations. Thismethodology was applied to a realistic Space Station assemblyoperation and compared with other alternatives. The results indicatethat the optimized attitude command trajectory results inthe smallest peak CMG momentum cost.Bernstein, J.J.; Xu, B.M.; Ye, Y.H.; Cross, L.E.; Miller, R.Dielectric Hysteresis from Transverse Electric Fields in Lead Zirconate Titanate Thin FilmsApplied Physics Letters, V74, N23, pp. 3549-3551Excellent symmetric dielectric hysteresis is observed from lead zirconatetitanate (PZT) thin films using transverse electric fields drivenby interdigitated surface electrodes. The 1-µm-thick PZT filmswith a Zr/Ti ratio of 52/48 are prepared on ZrO 2 buffered, 4-indiameter silicon wafers with a thermally-grown SiO 2 layer. Boththe ZrO 2 buffer layer and PZT film are deposited by using a similarsol-gel processing. Remnant polarization of about 20 µC/cm 2with a coercive field less than 40 kV/cm is obtained as measuredusing a triangle wave at 50 Hz. Thicker films are being developed,and retention for the transversely polarized state is currentlyunder study. One of the objectives of this study is to develop alarge array of d 33 -driven unimorph-sensing elements for a highresolutionacoustic imaging system.Bernstein, J.; Miller, R.; Kelley, W.; Ward, P.Low-Noise MEMS Vibration Sensor for Geophysical ApplicationsJournal of Microelectromechanical Systems, Vol. 8, No. 4, December 1999, pp. 433-438The need exists for high-sensitivity, low-noise vibration sensorsfor various applications such as geophysical data collection, trackingvehicles, intrusion detectors, and underwater pressure gradientdetection. In general, these sensors differ from classicalaccelerometers in that they require no dc response, but must havea very low noise floor over a required bandwidth. Theory indicatesthat a capacitive micromachined silicon vibration sensorcan have a noise floor on the order of 100 nano g/ over a 1-kHzbandwidth while reducing size and weight tenfold comparedwith existing magnetic geophones. With early prototypes, wehave demonstrated a Brownian-limited noise floor at 1.0 g/,orders of magnitude more sensitive than surface micromachineddevices such as the industry standard ADXL05.1999 Published Papers 79

Billingsley, G.O.; Kuchar, J.K.; Jacobson, S.W.Head-Up Display Symbology for Ground Collision AvoidanceGateway to the New Millennium; Proceedings of the 18th Digital Avionics SystemsConference (DASC), Saint Louis, MO, October 24-29, 1999, Vol. 1 (A00-21178 04-01),Piscataway, NJ, Institute of Electrical and Electronic Engineers, Inc., 1999, pp. 4.D.1-1 to4.D.1-8Four ground collision avoidance displays were tested using afixed-base T-38 simulator with a projection screen and simulatedHead-Up Display (HUD). When given a standard Break-X, pilotswere able to spend only 40 percent of the flight time betweendesired altitudes and crashed in 20 percent of the runs.Horizontally- and vertically-moving chevron symbols allowed 70and 80 percent of the flight time to be spent at the desired altitude,respectively, and resulted in a crash in 8 percent of the runs.A preview depiction using a perspective elevated surface at thedesired attitude was the best display for the task investigated,allowing 90 percent of the time to be spent at the desired altitudewith a crash rate of 2 percent.Boccuzzi, R.; Brown, T.; Cook, B.; Dodds, L.; Kochocki, J.; LeBlanc, M.; Robillard,M.; Stadelmann, E.; Stewart, W.; O'Brien, W.; Hudson, P.; Bracewell, T.; Farmer,K.; Gaborno, N.; Kono, K.; Vassar, E.A Simulation-Based Test and Evaluation Capability55th Institute of Navigation, Annual Meeting, Cambridge, MA, June 28-30, 1999,Proceedings (A00-18180 03-32), Alexandria, VA, Institute of Navigation, 1999, pp. 567-572A Simulation-Based Test and Evaluation Capability (SiBaTEC) providesa user with the ability to perform real-time Hardware-in-the-Loop (HIL) simulations. For a system under investigation, SiBaTECpermits design verification, testing of potential components, andsubsystem modifications before commitment to a prototype, andtesting of modified prototypes in simulation. Furthermore,SiBaTEC provides a means of performing system surveillancethrough repeatable monitoring and margin testing of hardwareand embedded software, and hypothesis testing of potentialaging and wear-out phenomena. SiBaTEC accomplishes theseactivities by means of a real-time simulation host supported by anetwork-based suite of tools and custom I/O capabilities. Thispaper describes these capabilities of SiBaTEC, as well as the currentsystem integration testing on a MK 6 guidance system.Boelitz, F.W.Kistler Launch Assist Platform (LAP) Return Burn ControlGuidance, Navigation, and Control Conference, Portland, OR. Sponsored by: AIAA, pp.1289-1299The Thrust Vector Control (TVC) design for the Return Burn segmentof the Kistler K-1 Launch Assist Platform (LAP) is presented.The design features a two-mode controller that initially providesstate-dependent pre-ignition orientation of the rocket engine,and on ignition, rapid orientation of the vehicle x-axis backtoward the launch zone. Following this pitch reversal, the controllerseamlessly switches to a second mode that features integralcontrol with acceleration direction estimation. This secondmode provides the fine pointing required by guidance to ballisticallyloft the LAP back to the launch zone. Control gains for thedesign are precomputed prior to launch through an automateddesign procedure that searches over a broad family of gains. Theautomated design tool simultaneously applies frequency domainand time domain constraints, which results in a controller thatachieves stable response with adequate margin and minimal settlingtime.Borenstein, J.T.; Gerrish, N.D.; Currie, M.T.; Fitzgerald, E.A.New Ultra-Hard Etch-Stop Layer for High-Precision Micromachining12th IEEE International Conference on Microelectromechanical Systems (MEMS), Orlando,FL, 1999, pp. 205-210In the current work, we describe a high-precision fabrication methodfor silicon micromachining based on a newly developed epitaxialetch-stop. This etch-stop, composed of a silicon-germanium alloywith no boron doping, outperforms traditional boron-doped etchstops in several important and fundamental ways. Etch selectivitiesin a variety of standard etchants compare favorably withthose obtained using high-concentration boron-diffused and epitaxiallayers. Microstructural analysis of the new etch-stop layerdemonstrates a significant reduction in defect density relative toboron-doped counterparts. Tuning-fork gyroscopes built with thenew etch-stop show build dimensions comparable to those fabricatedwith conventional methods. We propose a band structuremodel for the etch-stop mechanism that mimics the hole-injectionphenomenon often invoked for boron doping, and conclude witha brief discussion of the advantages of this new fabrication technology.Cantwell, R.H.; Ventresca, R.GPS Continuous Track on a Spinning Vehicle with Multiple Patch AntennaInternational Technical Meeting of the Satellite Division of the Institute of Navigation(ION GPS), Nashville, TN. Sponsored by: IONThe objective is to use multiple patch antennas to allow visibilityto the Global Positioning System (GPS) satellite vehicles for continuoustracking on a spinning platform without going throughthe acquisition process. Our methodology is to use patch antennasmounted to allow overlapped visibility coverage of the GPSsatellite vehicles. An inertial measurement unit is used to determinethe attitude and position or coverage of each antenna.Ephemeris is collected from each visible satellite. Using the GPSreceiver’s position and the positions of the GPS Satellite Vehicles(SVs), the receiver determines which satellites are in the field ofview of each antenna. The overlapped coverage allows an SV tobe acquired and tracked on two antennas. At this time, we do adirect track hand-over from the receiver channel whose antennawill be going out of view to another receiver channel using theantenna that has come into view. This eliminates interruption insignal tracking, and thereby results in continuous, accurate navigationsolutions. This technique has been demonstrated successfullyon multiple spinning vehicles.Cefola, P.J.; Nazarenko, A.I.; Yurasov, V.Refinement of Satellite Ballistic Factors for the Estimation of Atmosphere DensityVariations and Improved LEO Orbit PredictionSpace Flight Mechanics Meeting, Breckenridge, CO. Sponsored by: AAS/AIAAIn an earlier work, Prof. Nazarenko discussed an atmosphere densitytracking process that operates in parallel to the orbit determinationprocess. This atmosphere density tracking processemploys ballistic coefficient data observed over short arcs frommultiple satellites. The process includes: (1) a procedure for constructingthe density variations that operates on a 2- or 3-h grid,(2) a procedure for estimating the true ballistic coefficients of theemployed satellites that operates on a 28- or 56-day interval (1 or2 monthly solar cycles), (3) a procedure for forecasting the atmospheredensity at future times. This paper focuses on improvingthe algorithm for estimating the true ballistic coefficient of theemployed satellites. The main aspect of this improvement consistsof applying, for updating the ballistic factors of nonstandardsatellites, a linear function of altitude to model the systematic801999 Published Papers

errors. Numerical testing based on simulated data has beenundertaken to verify the correctness of the algorithm. The productsof the study include a proposal to work with the real data.The possibility of complex utilization of data from both NorthAmerican Defense (NORAD) and the Russian Space SurveillanceSystem is discussed.Chaudhry, A.I.; Thele, J.D.; Kang, D.S.High-Velocity Tele-Operated Rover13th Aerospace Defense Sensing Simulation and Controls (AeroSense), Orlando, FL.Sponsored by: SPIEThe High-Velocity Tele-operated Rover (HVTR) is motivated by agoal to exceed human physical speed with small ground vehiclesfor operations in an urban environment. A typical small (manpackable)ground vehicle’s speed tops out at 1-2 m/s (2-4 mph).Limited speed is attributed to real-time sensing and processing ofthe external environment. Low speed makes traversing multiplecity blocks taxing on the patience of a human operator.Traversing around a block may take 10-20 min. Even with operatorassistance, using video does not significantly increase thespeed due to the low perspective of the camera view and cameravibration in an outdoor setting.Connelly, J.; Kourepenis, A.; Larsen, D.; Marinis, T.F.Inertial MEMS Development for Space2nd International Conference on Integrated Micro Nanotechnology for SpaceApplications, Pasadena, CAMicromachined silicon inertial sensors offer revolutionaryimprovements in cost, size, and reliability for guidance, navigation,and control. Inertial sensors represent an important segmentof an emerging Microelectromechanical Systems (MEMS)technology, which combines semiconductor materials and processingto create integrated mechanical and electrical systems.Batch manufacturing techniques produce thousands of virtuallyidentical MEMS devices, each a few square millimeters in size,enabling inertial systems at a fraction of the cost, size, and powerof any previous technology. Development of MEMS inertialinstruments is driven by the high-volume, commercial marketthat targets modest performance applications at prices below$20 per axis. However, Draper has developed higher-performance,multi-axis systems using commercial processes to ensuretheir availability and affordability for lower-volume military andspace applications. The performance of these new MEMS inertialsystems is quickly approaching bias stability of 1 deg/h and 100µg and scale-factor stability of 100 ppm over -40°C to +85°C.Radiation testing is now underway to evaluate response to predictedspace environments. Future MEMS inertial systems willreflect a radical departure from the ways they have been conceived,fabricated, and tested in the past. New inertial deviceshave been incorporated, enabling multi-axis measurement in aplanar array, and development is underway on a new wafer-scaleprocess integrating sensors and Application-Specific IntegratedCircuits (ASICs) to create complete systems on a chip. These higher-performance,lower-power, inertial microsystems will be ideallysuited for many space applications. This paper addressesDraper’s inertial MEMS designs, fabrication methods, instrumentand performance progression, and development activities relatedto space applications. Space radiation issues for MEMS are discussed,expected environments are identified, and radiation testingof MEMS instruments is described. In addition, MEMSpackaging development toward high-level multi-axis system integrationis reviewed.Cunningham, B.T.; Regan, R.; Clapp, C.; Hildebrant, E.; Weinberg, M.; Williams, J.Miniature Silicon Electronic Biological Assay Chip and Applications for Rapid BattlefieldDiagnosticsBattlefield Biomedical Technologies, Orlando, FL, April 6, 1999. Sponsored by: SPIE. pp.26-34; 13th Aerospace Defense Sensing Simulation and Controls (AeroSense), Orlando,FL. Sponsored by: SPIEAssessing the medical condition of battlefield personnel requiresthe development of rapid, portable biological diagnostic assaysfor a wide variety of antigens and enzymes. Ideally, such an assaywould be inexpensive, small, and require no added reagents,while maintaining the sensitivity and accuracy of laboratorybasedassays. In this work, a MEMS-based biological assay sensoris presented that is expected to meet these requirements. Thesensor is a thin Silicon Membrane Resonator (SMR) that registersa decrease in resonant frequency when mass is absorbed onto itssurface. By coating the sensor surface with a monolayer of antibody,for example, we have detected the corresponding antigenwith a detection resolution of 0.25 ng/ml in a phosphate buffersolution. Micromachining techniques are being used to integratemany (64 elements on the first test chip) identical SMR sensorsinto a single silicon chip that would be capable of simultaneouslyperforming a wide variety of biomedical assays. The sensorsrequire only a small printed circuit board and an 8-V power supplyto operate and provide a readout. The presentation willdescribe the operation of the SMR sensor, the fabrication of thesensor array, and initial test results using commercially-availableanimal immunoglobulins in laboratory-prepared test solutions.de Fazio, T.L.; Rhee, S.J.; Whitney, D.E.Design-Specific Approach to Design for Assembly (DFA) for Complex MechanicalAssembliesIEEE Transactions on Robotics and Automation, Vol. 15, No. 5, pp. 869-81This paper uses Assembly Sequence Analysis (ASA) to exploreDesign for Assembly (DFA), subassembly partitioning, and assemblysequence choice for two complex assemblies. Complexassemblies have very high parts counts, a final assembly organizedas an assembly of subassemblies, and offer limited redesignoptions. ASA addresses combinatorial aspects of complex assembliesthat conventional DFA ignores: choice and partitioning ofsubassemblies and assembly sequence choice. This paperdescribes criterion-based searches for favorable subassembly partitioningand assembly sequences that use genetic algorithmtechniques to spread assembly move difficulty across entire finalassembly sequences while satisfying all logical constraintsimposed on the assembly sequence by part geometry. The measureof assembly move difficulty, a count of kinematic degrees offreedom secured during each final assembly step, is measured onan absolute scale. We find that ASA can pinpoint candidate DFArelatedredesigns and can suggest assembly issues to designers.Logical assembly issues dominate quantitatively characterizedissues when selecting assembly sequence or subassembly partitioning.After logical issues are addressed, the sequence choicecriterion defined here often duplicates choices made by experiencedanalysts. Finally, the sequence choice criterion favors inlineover final assembly lines.1999 Published Papers 81

Draim, J.E.; Cefola, P.; Proulx, R.; Larsen, D.; Granholm, G.R.Elliptical Sun-Synchronous Orbits with Line of Apsides Lying In or Near the EquatorialPlaneAstrodynamics Specialist Conference, Anchorage, AK. Sponsored by: AAS/AIAAThis paper explores the characteristics of retrograde, sunsynchronouselliptic orbits with line of apsides lying in or near theequatorial plane. Coverage plots for a five-satellite ring showingthe number of satellites in view and elevation angle data versuslatitude and local time are presented. Stability of the orbit is discussed.Also analyzed is the effect of the trapped radiation fieldenvironment (Van Allen Belts) on these orbits, as well as the exposureto damage by natural and man-made debris. A major advantageseen for these orbits is that they can be used to provideaugmented earth coverage for a selected latitudinal zone and aselected time of day (for all longitudes). This feature should proveuseful for nongeostationary satellite communications systemswhere increased capacity is needed during daytime peak-traffichours in heavily populated latitude bands.D'Souza, C.; Bogner, A.J.; Brand, T.; Tsukui, J.; Koyama, H.; Nakamura, T.An Evaluation of the GPS Relative Navigation System for ETS VII and HTV22nd AAS Guidance and Control Conference, Breckenridge, CO. Sponsored by: AASThe Global Positioning System (GPS) is being used increasingly forspacecraft navigation. Not only is GPS being used in the traditionalrole of absolute navigation, but it is also playing a role in relativenavigation, particularly for spacecraft rendezvous. Two suchinstances in which relative GPS navigation is playing a key role inspacecraft rendezvous are the Engineering Test Satellite 7 (ETS-VII) and the HII Transfer Vehicle (HTV). ETS-VII is a test satellitedeveloped by the National Space Development Agency of Japan(NASDA) and is designed to test the performance of a relative GPSsystem. HTV is the Japanese resupply vehicle for the InternationalSpace Station (ISS) and is being developed by NASDA. As currentlyenvisioned for HTV, the ISS will send its GPS measurementinformation over a Radio Frequency (RF) link to HTV, which willsimultaneously take GPS measurements to the same satellites.The HTV will difference the GPS measurements in a filter to providehighly accurate relative position and velocity information.NASDA has selected relative GPS navigation to be used for therendezvous from approximately 23 km from the ISS to 500 m,after which a laser sensor will be used to position the visiting vehiclefor grappling by an ISS arm. Draper Laboratory has evaluatedthe GPS relative navigation system for both ETS-VII and HTV. Thispaper will describe the testing methodology of the GPS relativenavigation system that was used to confirm the tests carried outby Mitsubishi Electric Corporation (MELCO) under contract toNASDA. The testing methodology used by Draper involved theuse of an RF satellite signal simulator. In addition to providing adescription of the filter architecture, MELCO also provided the trajectorydata to drive the satellite signal simulator. A NorthernTelecommunications (NorTel) Satellite Signal Simulator (SSS) usedthe target and chaser vehicle trajectory information to create theRF signal that a GPS receiver would expect to experience alongthe trajectory. Measurement data, including pseudorange anddelta range measurements, were recorded from the GPS receiver,with each trajectory being run separately. The recorded data fromthe two trajectories was then processed in a filter. Proper mergingof the measurement data in the relative navigation filterinvolved the synchronization of data from the two receivers. Onlythose measurements that were from common satellites wereused in the filter. The measurement data from the target and thechaser GPS receiver were time-tagged with slightly differenttimes. Therefore, in order to difference the measurements, theyhad to be brought to a common time. This was performed usinglinear interpolation, with the target measurement being the referencetime. The filter was an 8-state linearized Kalman filter. Thestates included the three relative position states, three relativevelocity states, a relative clock bias state, and a relative clock driftstate. The position and velocity states were expressed in the Hillframe, which is a curvilinear, rotating frame. The filter dynamicsfor the position and velocity were described by the well-knownHill-Clohessy-Wiltshire equations. An evaluation was also performedas to whether using the position and velocity for the targetand the chaser for the propagation would improve thenavigation results, and it was found to improve the results.However, the Hill-Clohessy-Wiltshire equations, which yield aclosed-form State Transition Matrix (STM), were still used in thepropagation of the covariance matrix. With the use of numericalintegration for the state equations and the state transition matrixfor the propagation of the covariance matrix, excellent filter performancewas obtained. The relative position accuracy was betterthan 0.4 m and the relative velocity accuracy was better than 6cm/s. During coasting periods of the trajectory, the velocity accuracywas better than 1 cm/s.Elwell, J.Inertial Navigation for the Urban WarriorThe International Society for Optical Engineering Conference, Proceedings, SPIE -International Society of Optical Engineering (USA), Vol. 3709, pp. 196-204Individual soldier geolocation in situations such as urban warfarewhere loss of Globl Positioning System (GPS) track can impact missionsuccess has become a critical problem. Concepts such as RF"time difference of arrival" and "dead reckoning" techniques havenot demonstrated their ability to support navigation reliablyinside buildings on their own. Inertial navigation is the only technologythat operates independent of external assets. The adventof micromechanical inertial sensor technology has resulted inlow-cost, very small, low-power navigation systems capable of fittingin a soldier’s boot. A miniature navigator consisting of threemicromechanical gyroscope and accelerometer packages, includingsupporting application-specific integrated circuit chips, andcapable of operating in support of such a mission has been developed.However, because of accelerometer and gyroscope drift,navigating inertially over long time periods, using even the mostprecise and most expensive inertial sensors available todayremains close to impossible. Inertial augmentation techniquesare therefore required, and the concept of personal inertial navigationsystems aided by zero velocity updating of the accelerometerswith each footfall has been examined and shown to besufficient to determine the location of an individual soldier accuratelywithin a large building complex after hours of operation. Inaddition to the accelerometer, updates of the gyro via zero attituderate techniques also enhance position accuracy, as well asprovide an attitude reference in support of soldier-carried targetingsensors.Faiz, R.I.Net RTM Preforming Process for Cost-Effective Manufacturing of Military Ground VehicleComposite Structures"Resin Transfer Molding," SAMPE Monograph, No. 3, pp. 127-138While the strength, stiffness, and signature advantages of compositematerials vs metals are well documented, their applicationto many military systems has been inhibited by the extreme costperformanceparadigm that has evolved in the industry, the poles821999 Published Papers

of which are high performance and price-insensitive aerospacecomponents and low-performance and price-sensitive automotivecomponents. Among the key requirements for military automotivecomposite parts are reasonably high fiber volumes(>40%) using a continuous reinforcement oriented in the productionvolumes at or less than the 10,000 level. A process thatpotentially can satisfy the needs of military automotive compositesproduction is Resin Transfer Molding (RTM). Cost studies inthe aerospace and automotive industries, and verified by Draper,have indicated that RTM holds great promise for reducing manyof the significant cost factors in composites manufacture, especiallyif the high cost of preform preparation can be reduced. Thispaper documents the successful definition and demonstration ofthe enabling technologies used for the development of costeffectivecontinuous reinforcement, oriented fiber, RTM preformssatisfying all performance requirements.Feder, H.J.S.; Leonard, J.J.; Smith, C.M. (reprinted)Adaptive Mobile Robot Navigation and MappingInternational Journal of Robotic Research, Vol. 18, No. 7, July 1999, pp. 650-668The task of building a map of an unknown environment and concurrentlyusing that map to navigate is a central problem inmobile robotics research. This paper addresses the problem ofhow to perform Concurrent Mapping and Localization (CML)adaptively using sonar. Stochastic mapping is a feature-basedapproach to CML that generalizes the extended Kalman filter toincorporate vehicle localization and environmental mapping. Theauthors describe an implementation of stochastic mapping thatuses a delayed nearest neighbor data association strategy to initializenew features into the map, match measurements to mapfeatures, and delete out-of-date features. The authors introduce ametric for adaptive sensing that is defined in terms of Fisher informationand represents the sum of the areas of the error ellipses ofthe vehicle and feature estimates in the map. Predicted sensorreadings and expected dead-reckoning errors are used to estimatethe metric for each potential action of the robot, and theaction that yields the lowest cost (i.e., the maximum information)is selected. This technique is demonstrated via simulations, in-airsonar experiments, and underwater sonar experiments. Resultsare shown for (1) adaptive control of motion and (2) adaptive controlof motion and scanning. The vehicle tends to explore selectivelydifferent objects in the environment. The performance ofthis adaptive algorithm is shown to be superior to straight-linemotion and random motion.Flueckiger, K.; Dowdle, J.A High Antijam INS GPS NavigatorAssociation of Old Crows, Adelphi, MD. Sponsored by: NAVWAR, April 7-8, 1999Traditionally, integrated inertial and Global Positioning System(GPS) sensing has been used to provide accurate high-bandwidthnavigation solutions. Algorithms designed to integrate these sensorshave not used the full sensor data available in a centralizedmanner. The so-called loosely coupled integration approachassumes that GPS provides a (low-bandwidth) Position, Velocity,and Time (PVT) solution to the navigation algorithm.Traditionally,tight integration approaches demand that GPS provide pseudorangeand delta-range measurements. In contrast, the DeepIntegration Algorithm, introduced here, uses raw in-phase andquadrature (I and Q) components from the GPS receiver’s correlatoroutputs. By using this full information from the receiver hardware,analysis and hardware results indicate that the DeepIntegration Algorithm will improve GPS jamming immunity significantly.A preliminary implementation of the Deep IntegrationAlgorithm for a single SV has been successfully embedded withina commercial-off-the-shelf C/A receiver. Results indicate that thereceiver loss-of-lock threshold can be extended by approximately15 to 20 dB in a sustained jamming environment. Results fromtwo dynamic scenarios are presented here: (1) a velocity stepalong the SV line-of-sight, and (2) a tactical munition scenario.Both scenarios are presented under a variety of jamming environments.The results are extrapolated to predict the performance ofthe Deep Integration Algorithm, with full multi-SV tracking capability,using P(Y)-code receiver hardware. This analysis is consistentwith performance predictions based on (software-only)simulation.Fuhry, D.Adaptive Atmospheric Reentry Guidance for the Kistler K-1 Orbital VehicleGuidance, Navigation, and Control Conference, Portland, OR. Sponsored by: AIAA, pp.1275-1288The Kistler K-1 is designed to be a fully reusable, two-stage launchvehicle for the economical delivery of small satellite payloads tolow earth orbit. Greatest efficiency and, hence, lowest cost areachieved by flyback of both vehicle stages to the near vicinity ofthe launch site. After deploying the payload and performing thenecessary phasing maneuvers, the second stage Orbital Vehicle(OV) performs a deorbit burn to achieve the desired trajectoryconditions at atmospheric entry. After entering the atmosphere,the OV is steered aerodynamically until it reaches the deploymentpoint for a stabilization parachute. Subsequently, drogue andmain parachutes complete vehicle deceleration for landing onairbags. This paper presents the design of an atmospheric guidancealgorithm for bank-to-turn steering of the OV prior todeployment of the stabilization parachute. Reentry guidance targetsa desired geographic position for deployment of the drogueparachute. The algorithm employs a numerical predictor/correctortechnique to compute the bank angle and the start time of asingle bank reversal required to null the predicted target positionmiss. Aerodynamic loads and heating are limited implicitly byselection of deorbit target conditions for reentry trajectory shaping.Results obtained using the Draper K-1 Integrated VehicleSimulation illustrates guidance performance under nominal anddispersed conditions.Granholm, G.R.; Proulx, R.J.; Cefola, P.J.Orbit Determination for Medium Altitude Orbits Using GPS Receivers and Ground-BasedTrackingAstrodynamics Specialist Conference, Anchorage, AK. Sponsored by: AAS/AIAAThe past few years have seen a proliferation of nontraditionalmedium or high-altitude constellations designed for use in communications.A constellation considered for use in the Ellipso systemfeatures ten satellites in highly elliptical Sun-SynchronousFrozen Line of Apsides (SSFLA) orbits. These orbits pose uniquechallenges to the orbit determination process, including higheccentricity, tesseral resonance, critical inclination, and sensitivityto solar radiation pressure. This paper will compare the effectivenessof Global Positioning System (GPS)-based tracking withground-based tracking in terms of accuracy of DifferentialCorrection (DC) solutions. A "Truth" orbit will be simulated using ahigh-precision Cowell numerical integrator.This orbit is then usedto create simulated GPS pseudoranges and ground-basedDoppler range and range-rate measurements. The GPS constellationis modeled and propagated using analytic J 2 -only equationsexpressed in equinoctial elements. To improve speed and performance,the GPS simulation is coded using a Message-PassingInterface (MPI) parallel implementation. The pseudoranges and1999 Published Papers 83

ange/range-rate observations are used in a least-square DCprocess to solve for state parameters, solar radiation and dragcoefficients, and ground station biases. Differences between thetouch and fit/predict orbits are analyzed numerically and graphically.Cases are run for both atmospherically quiet and perturbedepochs and with atmospheric and gravitational mismodeling. It isfound that the accuracy of the solution is strongly affected byatmospheric conditions. Both methods yield similar solutions, butthe DC process scan requires more ground-based observationsthan GPS pseudoranges. Overall, both tracking methods areshown to be viable for these types of orbits.Guinon, W.; Setterlund, R.H.; Phillips, R.Reducing the Power Requirements of an Interferometric GPS Receiver for SpacecraftAttitude DeterminationVision 2010: Present and Future: National Technical Meeting, San Diego, CA. Sponsoredby: ION, pp. 561-573The ongoing development of micromechanical inertial systemsthat require very little power suggests the concomitant developmentof a low-power Global Positioning System (GPS) receiver.The combination of such an interferometric receiver with inertialinstruments would fill the need for a low-cost, lightweight, lowpowerattitude determination system for use in small, low-costsatellites with modest accuracy requirements (0.1 to 0.5 deg).Power consumption by the GPS receiver can be reduced by turningoff the RF front end, the frequency synthesizer, the referenceoscillator, and the digitizer for brief intervals of time while usingthe inertial system to maintain adequate attitude knowledge andto simplify obtaining subsequent IGPS attitude updates withouttime-consuming integer ambiguity resolution. This study looks atthe implications of this strategy on the details of the receiveroperation and design including reacquisition, and the tradebetween pre- and post-detection integration, as well as powerconsumption. The accuracy of such a system as a function of theinterval between GPS measurements was assessed. Dependingon this interval and on other parameters, such as IMU quality,antenna baseline, etc., system power consumption on the order of1 W or less can be achieved. Accuracies in the 0.1- to 0.5-degregime are readily achievable. Volume, weight, and power projectionsare based on existing technology and hardware, leading toa system concept for a spacecraft attitude determination thatcould be of enormous benefit for small satellites.Gustafson, D.E.GPS Signal Tracking Using Maximum-Likelihood Parameter EstimationJournal of the Institute of Navigation, Vol. 45, No. 4, Winter 1998-1999, pp. 287-295This paper considers the problem of Global Positioning System(GPS) carrier tracking in the presence of spurious modulationcomponents in amplitude and phase caused by rotating antennaelements in a reentry body. The commonly used phase-lock loopapproach is not adequate since the disturbances are not modeledspecifically. This is a nonlinear estimation problem that isattacked here as a linear estimation problem with an unknownmodulation parameter. The modulation parameter is estimatedusing a maximum-likelihood method in an architecture that usestwo Kalman filters of similar structure, one for parameter estimationand one for state estimation. This architecture uncouples thestate and parameter estimation processes and reduces the tendencyto build up incorrect correlations in the estimator. The performanceof the estimator is studied using a Monte Carlo simulation.Maximum likelihood results are found to be superior to thoseobtained using a second-order phase-lock loop and a nonadaptivefourth-order Kalman filter.Haley, J.F.; Hanson, D.S.; Marinis, T.F.Integration of Resistors and Capacitors within MCM-L SubstratesIMAPS (International Microelectronics and Packaging Society) ATW (Advanced TechnologyWorkshop) on Integrated Passives, Denver, CO, April 1999, and IMAPS New EnglandChapter Annual Symposium, May 1999The ever-increasing complexity of microelectronics has oftenresulted in greater overall packaging volumes. Although theMultichip Module-Laminate (MCM-L) has been successful inreducing volume by bringing several types of componentstogether and connecting them on substrates of numerous signallayers, the next step should involve the integration of theseadvantages. The focus of this research has been on layering passivecomponents within MCM-L substrates. Processes have beendeveloped to integrate two kinds of passive components: resistorsand capacitors into the laminate structure of the MCM-L toincrease its density and thereby decrease its volume. The resistorintegration process has successfully interconnected componentsembedded through lamination between nonreinforced epoxyresin and copper foil. The capacitor integration process has successfullyinterconnected components embedded through laminationbetween nonreinforced epoxy resin and copper foil. Thecapacitor integration process involved the development of both amaterial of high dielectric constant, and a means for laminatingthis material between copper to create very thin parallel platecapacitors. The dielectric material developed consisted of a lowviscosity thermoset polymer mixed with a high-permittivityceramic powder. A detailed description of each of the twoprocesses is presented, including difficulties encountered in interconnectingthe embedded resistors and mixing phenomenaobserved for the polymer/ceramic dielectric composite.Hammett, R.C.Ultra-Reliable Real-Time Control Systems – Future TrendsIEEE Aerospace and Electronic Systems Magazine, Vol. 14, No. 8, pp. 31-36Today’s aircraft use ultra-reliable real-time controls for demandingfunctions such as Fly-by-Wire (FBW) flight control. Future aircraft,spacecraft, and other vehicles will require greater use ofthese types of controls for functions that currently are allowed tofail, fail to degraded operation, or require human intervention inresponse to failure. Fully automated and autonomous functionswill require ultra-reliable control. But ultra-reliable systems arevery expensive to design and require large amounts of onboardequipment. This paper will discuss how the use of low-cost sensorswith digital outputs, digitally commanded fault-tolerantactuation devices, and interconnecting networks of low-cost databuses offer the promise of more affordable ultra-reliable systems.Specific technologies and concepts to be discussed include lowcostautomotive and industrial data buses, "smart" actuationdevices with integral fault-masking capabilities, management ofredundant sensors, and the fault detection and diagnosis of thedata network. The advantages of integrating the control and distributionof electrical power with the control system will be illustrated.The design, installation, and upgrade flexibility benefitsprovided by an all-digital and shared network approach will be841999 Published Papers

presented.The economic benefits of systems that can operate followingfailure and without immediate repair will be reviewed. Theinherent ability of these redundant systems to provide effectivebuilt-in test and self-diagnostics capabilities will be described.The challenges associated with developing ultra-reliable softwarefor these systems and the difficulties associated with exhaustiveverification testing will be presented, as will additional developmenthurdles that must be overcome.Hattis, P.; Bailey, R.Overview of the Kistler K-1 Guidance and Control SystemGuidance, Navigation, and Control Conference, Portland, OR. Sponsored by: AIAA. pp.1247-1254Draper Laboratory, under contract to Kistler AerospaceCorporation, is developing the complete flight guidance and controlsoftware for the K-1 launch vehicle. The K-1 is a fully reusabletwo-stage vehicle.The entire K-1 flight is performed autonomouslyexcept for one uplink of expected landing site winds beforelanding. The Launch Assist Platform (LAP) first stage flies a nearlyopen-loop three NK-33 engine boost phase, and subsequentlyreignites the center engine after staging to enable return to thelaunch site in a controlled coasting flight. The Orbital Vehicle (OV)second stage uses a single NK-43 engine to fly a fully closed-loopascent to orbit. The LAP trajectory is designed to trade return propellantrequirements against OV ascent performance impacts.Orbital maneuvers that are computed onboard the OV are used tocircularize its orbit, deploy the payload, rephase the OV orbit forlanding after 24 hours, and then deorbit the vehicle. The OV reentryis flown with control thrusters used to bank the vehicle asdirected by a predictor-corrector guidance law. Landing of bothstages is done with a parachute descent and airbag touchdown,with return trajectories biased to provide correction of expectedparachute wind drift effects during descent.Henderson, T.; Dennehy, N.Attitude Control and Energy Storage (ACES) Flywheel Demonstration Testbed17th Space Power Workshop, Long Beach, CANO ABSTRACTHouston, K.M.; Hillman, R.E.; Kobler, J.B.; Meltzner, G.S.Development of Sound Source Components for a New Electrolarynx Speech Prosthesis24th International Conference on Acoustics, Speech and Signal Processing (ICASSP),Phoenix, AZ. Sponsored by: IEEE, March 15-19, 1999For many individuals who lose their voices due to laryngeal canceror trauma, the only option for speech is to use an Electrolarynx(EL), which is battery-powered vibrator that is held to the throat.Current devices produce speech that sounds very machine-likewith low levels of loudness and intelligibility, and which alsodraws undesired attention to the user. A project at Draper, theMassachusetts Eye and Ear Infirmary, and MIT aims to develop amuch improved EL called the Electrolarynx CommunicationSystem (ELCS), which is a DSP-based device consisting of soundsource, control, and speech enhancement subsystems or modules.This paper introduces the ELCS and discusses developmentsto date in the sound source module. Specific topics include thedesign of a new linear EL transducer and investigations into glottalwaveform synthesis that should result in much more naturalspeech output.Kabir, A.E.; Bashir, R.; Bernstein, J.; DeSantis, J.; Mathews, R.; O’Boyle, J.O.;Bracken, C. (reprinted)High-Sensitivity Acoustic Transducers with Thin p+ Membranes and Gold BackplateSensors and Actuators A-Physical, Vol. 78, No. 2-3, December 14, 1999, pp. 138-142High-sensitivity acoustic transducers (microphones) have beenfabricated on 5-in wafers in a production environment and theexperimental results are presented. One main advantage of thismicrophone design is that it can be fabricated on a single wafer,eliminating the need for the multiple wafers and subsequentwafer bonding steps as in conventional designs. The devices usethin (similar to 3 µm) p+ silicon membranes as the active movableelement and a thick perforated plated gold backplate. The p+membranes are fabricated using an optimized boron solid sourcediffusion at 1150°C. Ethylene-Diamine-Pyro-Catecol (EDP) etchingat 100°C was performed from the backside of double-sided polishedwafers to release the thin silicon membranes. The zero-biascapacitance with the air gap was 2.2 pF, and it increased to 2.4 pFat 9 V. The frequency response was measured, and the measuredsensitivity of 5.28 mV/Pa at 5 V and 10.77 mV/Pa at 9 V at 1 kHz areamong the highest reported in the literature for micromachinedacoustic transducers.Kirkos, G.A.; Jurgilewicz, R.P.; Duncan, S.J.MEMS Optimization Incorporating Genetic AlgorithmsThe International Society for Optical Engineering Conference, Proceedings of the SPIE -International Society of Optical Engineering, Vol. 3680, Pt. 1-2, pp. 84-93Micromechanical sensors are simulated routinely using finite-elementsoftware. Once a structure has been proposed, various parametersare optimized using experience, intuition, and trial anderror. However, using proven finite-element modeling coupledwith a Genetic Algorithm (GA), optimal designs can be "evolved"using a hands-free approach on a workstation. Once a problem isdefined, the sole task required of the designer is the specificationof a mathematical objective function expressing the desiredproperties of the sensor; the sensor geometry that maximizes thegiven function is then synthesized by the algorithm. We havedeveloped an optimization tool and have applied it to the designof Tuning-Fork Gyroscopes (TFGs). In this paper, we demonstratehow a TFG was optimized using GAs. TFG suspension beamlengths were adjusted through the robust search technique,which is resistant to trapping in local maxima. Desired vibrationmode order and mode frequency separations were governed bythe objective function as specified by the designer. This multidimensionalnonlinear optimization problem had a solutionspace of over 8 million possible designs. Industry-standardmechanical computer-aided engineering tools were integratedalong with a GA toolbox and a web-based control interface.Designs offering reduced vibration sensitivity and increased sensordynamic range have been produced. A tenfold decrease intotal sensor optimization time has been documented, resulting inreduced development time.Kogan, R.G.; Desai, M.; Pien, H.; Grimson, E.Model-Based Visualization of Ultrasound ImagesBattlefield Biomedical Technologies, Orlando, FL. Sponsored by: SPIE, pp. 84-92Ultrasound imaging is the most pervasive, cost-effective, portable,high-resolution, and nonionizing modality of diagnostic imagingavailable. The use of ultrasounds, however, has been hampered bythe noise properties and poor contrast inherent in such imagery.1999 Published Papers 85

A novel processing system is currently being developed thatovercomes some of these disadvantages by producing a highqualityrendering of the anatomical structure of interest. In particular,a normal anatomical atlas is used as the starting point; thisatlas is produced from either CT or MR imagery. As the ultrasoundprobe is moved along the body, image registration techniques, aswell as external instrumentation that monitors the position andattitude of the ultrasound probe, are used to provide a continuousmapping between the ultrasound observations and theatlas. As discrepancies between the atlas and the observed anatomyoccur, the atlas is deformed to reflect actual observations.Operated in this mode, the system displays the deformed highresolutionatlas to the user, providing a high-contrast, low-noiserendering of the patient’s anatomy. In scenarios such as battlefieldcritical care, where large, immobile CT or MR scanners are notfeasible, deformation of a high-quality atlas to match real-timeultrasound imagery can provide much improved assessment andtreatment possibilities.Kourepenis, A.Low-Cost MEMS Inertial Systems for GPS Antijam ApplicationsDraper Report, April 8, 1999Microelectromechanical System (MEMS) technologies have theenormous potential to enable the realization of low-cost inertialsystems for a myriad of both commercial and military applications.With the large volume needs of the commercial marketseeking inertial systems for automotive, camcorder, toys, andother applications, an economical base for the low-cost manufactureof these technologies will be established. These same capabilitiescan be leveraged to realize low-cost inertial systemscritical to the development and deployment of weapons platformsat costs that cannot be matched in other technologies.Current applications being demonstrated with MEMS inertialtechnologies include competent munitions, autonomous vehicles,robotics, and personal navigation. Many of these applicationsuse MEMS-based Inertial Navigation System (INS)/GlobalPositioning System (GPS) systems to enable precise guidance,navigation, and control functions at power, volume, and g-survivabilitylevels unattainable by other means. The current performanceof these devices is in the 10- to 100-deg/h range and offermarginal improvement in the Antijam (AJ) capability of GPSreceivers. Improvements in inertial performance, combined withinnovative techniques for coupling the inertial and GPS systems,will result in excellent rejection of intentional and unintentionalGPS interference at low cost. This presentation details the stateof-the-artin MEMS technologies, highlighting current levels ofperformance, and future initiatives that will result in small, lowcostinertial systems that perform at levels in the 0.01- to 0.1-deg/h regime. Operating principles and current levels ofperformance of Draper’s MEMS technologies will be presented,and various applications and demonstrations will be described.Initiatives and roadmaps to higher levels of performance and theapplicability to INS/GPS high AJ will be provided.Kwan, A.; Bedrossian, N.S.; Jang, J.W.; Grigoriadis, K.Reducing Conservatism of Analytic Transient Response Bounds via Shaping FiltersAstrodynamics Specialist Conference, Anchorage, AK. Sponsored by: AAS/AIAARecent results show that the peak transient response of a linearsystem to bounded energy inputs can be computed using theenergy-to-peak gain of the system. However, an analytically computedpeak response bound can be conservative for a class ofbounded energy signals, specifically, pulse trains generated fromjet firings encountered in space vehicles. In this paper, shaping filtersare proposed as a methodology to reduce the conservatismof peak response analytic bounds. This methodology was appliedto a realistic Space Station assembly operation subject to jet firings.The results indicate that shaping filters indeed reduce thepredicted peak response bounds.McConley, M.W.; Oh J.H.; Jamoom, M.B.; Feron, E.Solving Control Allocation Problems Using Semi-Definite ProgrammingJournal of Guidance Control and Dynamics, Vol. 22, No. 3, pp. 494-497We consider the control surface allocation problem in the casewhen the surface allocation is limited to be a linear mapping frommoment space to control space. We show that an approach tothat problem based on ellipsoid volume maximization can beeasily recast as a convex optimization problem. This method isapplied to a numerical model of the F-18 High-Alpha ResearchVehicle (HARV) and has been compared with other approaches.The convex nature of the optimization problem under considerationmakes it possible to incorporate the proposed procedure in areal-time aircraft control allocation reconfiguration in the event ofdamaged control surfaces. The byproducts of the optimizationprocedure (especially the resulting ellipsoids) may be used inother proposed surface allocation procedures as well.McGovern, L.K.; Feron, E.Closed-Loop Stability of Systems Driven by Real-Time, Dynamic Optimization Algorithms38th IEEE Conference on Decision and Control, Phoenix, AZ. Sponsored by: IEEEThe Receding Horizon Control (RHC) scheme uses on-line optimizationto find a finite-horizon control input to a constraineddynamic system. This paper examines the relationship betweenthe optimization algorithm and the closed-loop dynamic systemin RHC. Past research on RHC has assumed that the optimizationalgorithm provides an optimal solution in a fixed time interval.Since RHC typically employs quadratic programming, which isusually solved only approximately, this presupposition is not valid.Instead of making the traditional optimality assumption, thispaper supposed that the provided solutions are only suboptimal.A sufficient condition is derived for closed-loop stability givencontrol sequences, which are optimal with tolerance. Also, abound is derived for the number of computations to find an optimalsolution from a warm start using an interior-point method. Aslong as this number of computations can be carried out in lessthan the time step of the dynamic system, the closed loop is guaranteedto be stable.Miller, R.; Eiceman, E.; Alan, G.; Nazarov, E.A Micromachined Field Asymmetric-Ion Mobility Spectrometer (FA-IMS)8th International Conference on Ion Mobility Spectrometry, Buxton, Derbyshire, UKThe possibility of creating mobility spectrometers with dimensionsunder a few centimeters has been considered a next step inreducing the size and cost of Ion Mobility Spectrometry (IMS)analyzers. This would be plausible if the drift tube design couldbe simplified over traditional configurations and if fabricationmethods were amenable to mass production. One drift tubedesign, which is simplified without ion shutters or voltagedividers and offers extremely high sensitivity, is the asymmetricfield analyzer or FA-IMS. The FA-IMS has ion behavior rooted inmobility phenomena, although the details of ion motion areunder active investigation. This mobility analyzer offers novelty inion behavior, which is a secondary consideration, and was selectedfor the simplicity of a planar micromachined construction.Another factor was the potential for improved detection limits861999 Published Papers

over conventional analyzers. A planar micromachined field asymmetric-mobilitydrift tube has been crafted and partly characterized;results of these studies will be described along with certainperformance features. The micromachined spectrometer drifttube is ~2.5 x 2.5 x 0.2 cm 3 , and is equipped with a 10.6-eV(l=116.5 nm) photo-discharge lamp as the ion source. The flowrate of drift gas is 21 min (-1) of scrubbed air. Results from parametricstudies with organic vapors of environmental, medical, orsecurity interests are being used to evaluate the performance ofthe drift tube. Any compromises in spectral characteristics fromreduction in size are still being assessed, but do not appear to besignificant.These include detection limits, signal to noise, the effectof moisture, and resolution of ions. Ion identities were confirmedby interfacing the analyzer to a tandem mass spectrometer.Muldoon, R.C.; Gill, J.; Brock, L.D.Integrated Mechanical Diagnostic (IMD) Health and Usage Monitoring System (HUMS) -An Open System Implementation Case StudyGateway to the New Millennium; Proceedings of the18th Digital Avionics SystemsConference (DASC), Saint Louis, MO, October 24-29, 1999, Vol. 2 (A00-21262 04-01),Piscataway, NJ, Institute of Electrical and Electronic Engineers, Inc., 1999, pp. 9.B.4-1 to9.B.4-8.The large number of flight-critical and maintenance-intensivedynamic components in a helicopter has led to the developmentof diagnostic systems to improve flight safety and reduce operationsand support costs. These systems, commonly called Healthand Usage Monitoring Systems (HUMS), have been used on largecommercial helicopters and are now being developed for militaryhelicopters. In 1997, the U.S. Navy embarked on an innovativeapproach to fielding an integrated mechanical diagnostic systemfor Navy helicopters. Sponsored by the Department of Defense’sJoint Dual Use Program Office, the Navy and BF GoodrichAerospace began an accelerated program to field a militarizedversion of a commercial HUMS for the CH-53E and SH-60B helicopters.An open systems architecture is critical to the militaryimplementation of HUMS technology on multiple platforms. Anopen system architecture is essential for system upgrades and forintroducing the broad range of new technologies as they becomeavailable from a number of different sources. Through significantteam effort and participation from BF Goodrich, the Navy, DraperLaboratory, the Joint Advanced HUMS (JAHUMS) program team,and the DoD Open Systems Joint Task Force (OS-JTF), the IMD-HUMS program is successfully navigating the often-turbulentwaters of open system architecture implementation. This paperwill provide details on some of the lessons learned from both anindustry and government perspective and provide a program statusupdate.Nazarenko, A.I.; Cefola, P.; Proulx, R.J.; Yurasov, V.Neutral Atmosphere Density Monitoring Based on Space Surveillance System Orbital DataAstrodynamics Specialist Conference, Anchorage, AK, August 16-19, 1999. Sponsoredby: AAS/AIAAOne approach for increasing the accuracy of satellite orbit determinationand prediction for Low-Earth Orbit (LEO) satellites is theorganization of an upper atmosphere monitoring function. Thiswould be the analog of a weather service in the lower atmosphere.Monitoring the upper atmosphere based on the use of theavailable satellite atmospheric drag data (ballistic factors) on allcatalogued LEO satellites offers a low-cost approach to this capability.These data are operationally updated in the SpaceSurveillance System (SSS) as a result of regular satellite observations.It is concluded that there are actual possibilities for operationalmonitoring of the global atmospheric density variations ataltitudes ranging from 200 up to 600 km. The elaboration of aplan for a real data test of the upper atmospheric monitoring conceptis discussed.Persson, B.A.Control of the Kistler K-1 First-Stage Reorientation Prior to EntryGuidance, Navigation, and Control Conference, Portland, OR. Sponsored by: AIAA, pp.1300-1309The control design for the Kistler K-1 first-stage reorientationmaneuver is presented. The reorientation maneuver follows themain engine burn that places the first stage on a trajectory toreturn it to the landing site. The design takes advantage of theavailable aerodynamic moment to accelerate the first stage in thedirection of the desired orientation for entry. The maneuver takesplace near the apex of the first-stage return trajectory, a region oflow dynamic pressure. Small Attitude Control System (ACS)thrusters are used to stabilize the vehicle during the maneuver.The maneuver must therefore complete prior to reentry. Thispaper also presents a method for incorporating this time constraintinto the design.Proulx, R.J.; Smith, J.E.; Cefola, P.J.; Draim, J.E.Optimal Station-Keeping Strategies via Parallel Genetic AlgorithmsSpace Flight Mechanics Meeting, Breckenridge, CO. Sponsored by: AAS/AIAAIn an effort to overcome the limitations of more traditional methods,this paper investigates the use of genetic algorithms in generatingnongreedy, global, near-optimal station-keepingstrategies. The orbit of an Ellipso TM Borealis satellite is constrained,and the minimum-fuel optimal burn strategy is developed suchthat the orbit is maintained within the specified constraints overthe entire time period of interest. The resulting fuel costs areshown to be lower than costs estimated via previous methods,specifically previous primer vector strategies. Operational andcomputational limitations of this method are also described.Rubenstein, D.S.; Carter, D.W.Attitude Control System Design for Return of the Kistler K-1 Orbital VehicleJournal of Spacecraft and Rockets, Sponsored by: AIAAAn attitude control system design is presented that provides themaneuver capability and aerodynamic angle maintenance necessaryfor the atmospheric reentry and return to launch site of anunmanned reusable launch vehicle.The primary functions are categorizedinto those that perform bank maneuvers about the airrelativevelocity vector and those that are responsible for thetracking and control of the vehicle aerodynamic trim conditions.The control system is supported by an onboard aerodynamic estimationfunction. The estimator uses measurements of vehiclestates from navigation in combination with analytic models in again-scheduled filter environment to provide control with currenttrim angle information. The control system uses this informationto minimize actual vehicle deviations from the trim. Also, controlis provided with bank commands from a guidance function. Asthis paper is concerned only with the control and estimation functions,the guidance strategies are discussed only to the extentthat is necessary to justify/clarify control or estimator designs.The algorithms developed here are applied to the Kistler K-1Orbital Vehicle and tested in the Kistler Integrated VehicleSimulation at Draper. Results indicate that the approach toentry/return control is both fuel efficient and effective from alanding accuracy perspective.1999 Published Papers 87

Rubenstein, D.S.; Melton, R.G.Multiple Rigid-Body Reorientation Using Relative Motion with Constrained Final SystemConfigurationJournal of Guidance, Control, and Dynamics, Vol. 22, No. 3, 1999, pp. 441-446Movable appendages in multibody spacecraft can augment orreplace the attitude control actuators. In this work, motions of themovable bodies relative to the main body are used to adjust thesystem’s inertial attitude to approach or attain a desired targetattitude. A control algorithm designed to generate the maneuvercommands that cause the necessary relative motions is testedwith several cases representing a variety of dynamic conditions.The control can accommodate many different system configurationsand dynamic conditions, such as nonzero system momentum,a problem that historically has proved difficult to solve in ageneralized, three-dimensional mode. Additionally, the controlcan return the system’s geometric configuration to its initial stateby the conclusion of the reorientation. The results indicate thatthe control can accomplish nearly complete reorientations in allcases tested while meeting the system constraints.Sacramone, A.; Desai, M.Real-Time Detection of Undersea Mines. A Complete Screening and Acoustic FusionProcessing System13th Aerospace Defense Sensing Simulation and Controls (AeroSense), Orlando, FL, April5-9, 1999. Sponsored by: SPIE; 4th Detection and Remediation Technologies for Minesand Mine-like Targets, Orlando, FL, April 5-9, 1999. Sponsored by: SPIE, pp. 615-625A complete mine Detection/Classification (D/C) system has beenspecified and implemented that runs in real time and has beenexercised on the latest available dual-frequency side-scan sonaracoustic image sets. The complete D/C system comprises a collectionof algorithms that has been developed and evolved atDraper over the past decade. The detection process consists ofimage normalization, enhancement, segmentation (blob formation),and feature extraction algorithms. The enhancement algorithmis a variant of a Markov random field-based anomalyscreener developed in FY 94. The features that were extractedwere those derived in FY 93. A distance constrained matchingalgorithm, which was developed in FY 95, is used to generate a listof High- and Low-Frequency (HF and LF) fused tokens. The classificationprocess involves the evaluation of a hierarchy of three,multilayer perceptron neural networks: HF, LF, and HF/LF fused.Research performed in FY 95 also concentrated on the developmentof several variants of information fusion with a hierarchicalneural network. The "discriminant-combining" variant of fusionwas selected as part of this D/C system. In addition, a classificationpost-processing and decision node statistic modificationstep, which was developed in FY 96, was included. This paper willdescribe the algorithms that were implemented. However, theemphasis will be on the performance results of processing the latestavailable side-scan imagery, comparison of single sensor vsdual-frequency sensor result, and the issues that were encounteredwhile exercising the D/C system on the new data set.Scholten, J.R.; Burnes, J.R. III; Gels, R.G.; McKenna, J.F.; Rosenberg, S.C.;Rosenstrach, P.A.The Smart Intrusion Sensor AlarmDSP World ICSPAT, Orlando, FLThe Smart Intrusion Sensor Alarm (SISA) is a small (10-cm) batterypowereddevice with a miniature geophone and microphone sensor,flexible Digital Signal Processor (DSP)-based signal processing,and a short-range radio transmitter. Advanced processing andpackaging technologies are used to minimize size and maximizecapability. Prototypes have been fabricated and field-tested todetect motorized vehicles and footsteps and to trigger a remotecamera.The SISA will operate outdoors unattended for over 2 weeks.Schwartz, G.; Richter, D.A Concept for a Survivable Ship Control Computer12th Ship Control Systems Symposium, The Hague, Netherlands. Sponsored by: SCSA surface ship can withstand considerable physical damage, but ifthe ship’s control system does not also survive, the ship might stillfail to complete its mission, or worse. This is in contrast to an aircraftfly-by-wire control system, where a small amount of damagemay well cause loss of the vehicle, and the survivability of the controlsystem is less of an issue. Fault-tolerant computers for realtimeapplications have typically followed the aircraft modelwithout survivability as a goal. For example, the Ship ControlComputer for the Seawolf attack submarine, SSN-21, wasdesigned to be highly fault tolerant, but the redundant computingchannels are connected via dedicated communication linksand all are located in a single, relatively confined ship space. Evenminor damage in that space could cause the loss of all processingchannels. Furthermore, regaining capability would require a considerablerepair effort. A concept is presented for a control systemcomputer that will survive damage as well as tolerate faults. Theconcept has been pursued as an Independent Research andDevelopment project at Draper. The concept takes advantage ofthe redundant paths intrinsic to a mesh network. The computer’sredundant computing channels communicate with each otherusing a connection-oriented protocol over a mesh network,which is automatically reconfigured by the surviving channelsafter damage has occurred. Thus, a channel cut off by damage tothe network could be reconnected instantly, without the delayand expense of installing new cable. In the event of widespreaddamage, a channel could be plugged into the network in anundamaged part of the ship. The network technology beingemployed for this project is Asynchronous Transfer Mode, but theconcept could be realized with other technologies. It is envisionedthat when a fully developed computer is transitioned forshipboard deployment, the redundant computer channels will benodes on the ship’s data network infrastructure.Sitomer, J.; Connelly, J.; Kourepenis, A.Micromechanical Inertial Guidance, Navigation, and Control Systems in Gun-LaunchedProjectilesAtmospheric Flight Mechanics, Portland, OR. Sponsored by: AIAAMicromechanical technology applied to inertial instrumentsopens up many new applications where cost, size, and power areimportant. One very important application is the guidance, navigation,and control of gun-launched projectiles. In order to beaffordable, these systems must cost less than $2,000, have verylow power requirements, and eventually fit into a standard NATOfuze of 9 in 3 , including all fuzing and safe and arming functions.Since 1997, when Draper demonstrated the first successful launchfrom a Navy 5-in projectile, many new applications have beenidentified and are being pursued. Draper is currently ready toflight test a Micromechanical Inertial Measurement Unit/GlobalPositioning System (MMIMU/GPS) in a spinning 5-in projectilewith the system despun in the nose-mounted fuze assembly. Thispaper describes the application of this technology to Navy andArmy projectiles, both spin stabilized and nonspinning. Some ofthe projectiles described will be the Navy’s Extended-RangeGuided Munition (ERGM) Demo, the Army’s Precision GuidedMortor Munition (PGMM) 120-mm projectile, and standard NATOfuze applications to spin-stabilized projectiles, such as the existing5 in, 155 mm, and 120 mm, etc.881999 Published Papers

Smith, J.; Proulx, R.J.; Cefola, P.; Draim, J.E.An Operational Approach for Generating Near-Optimal Station-Keeping Strategies viaParallel Genetic AlgorithmsAstrodynamics Specialist Conference, Anchorage, AK. Sponsored by: AAS/AIAAExtending on the results of the authors’ previous parallel geneticalgorithm optimization approach, this study investigates ways inwhich the parallel genetic algorithm can be used as the basis foran operational station-keeping system. Specifically, an orbit isdefined and parallel genetic algorithms are applied in such amanner that the orbit is maintained within a given set of tolerances.However, unlike the previous study that focuses only onmaintaining the orbit within the state constraints, this study focuseson ways to maintain near optimality in the station-keepingmaneuvers, while also maintaining the operational characteristicsof repeatability, speed of convergence, and ease of implementation.Finally, the use of this operational station-keeping algorithmas a planning tool is discussed.Soltz, J.A.; D'Souza, C.; Brand, T.J.; Tsukui, J.; Koyama, H.; Nakamura, T.An Evaluation of the GPS Relative Navigation System for HTV Using a FunctionalSimulatorAIAA ISS Service Vehicle Conference, Houston, TX.No AbstractStaugler, A.; Shepperd, S.W.Autonomous On-Orbit Targeting and Guidance for a Reusable Launch VehicleGuidance, Navigation, and Control Conference, Portland, OR. Sponsored by: AIAA, pp.1255-1265Kistler Aerospace is currently developing a two-stage reusablelaunch vehicle to place payloads in circular low-Earth orbit. Thisvehicle is unique in that both stages return to the vicinity of thelaunch site on parachutes and land on airbags. Draper has developedthe guidance and targeting algorithms for the on-orbitphase of the Kistler mission. This phase consists of five burns, foreach of which targeting supplies the desired terminal conditionsby directing the thrust vector during the burn. Targeting is drivenby two primary requirements. First, the vehicle maneuver capabilityon entry is limited. Targeting accuracy is essential for landingin a small zone near the launch site. Second, unlike almost allprevious programs, the Kistler vehicle will not communicate withthe ground while on orbit, so targeting must be completelyautonomous. Guidance draws heavily on algorithm designs fromthe Space Shuttle, but a new guidance mode developed specificallyfor the Kistler mission to provide an accurate orbital periodfor phasing is described.Tetewsky, A.; Lozow, J.B.; Flueckiger, K.W.Determining Specifications for an External GPS Reference OscillatorInternational Technical Meeting of the Satellite Division of the Institute of Navigation(ION GPS), Nashville, TN. Sponsored by: ION, pp. 289-300There are many test and development situations in which aGlobal Positioning System (GPS) receiver is operated with anexternal frequency reference that is phase locked to a GPS simulatoror to a second receiver. Although one might intuitively predictthat receiver performance could be enhanced by using ahigh-quality commercial frequency source, the application of thissource can actually degrade performance as evidenced byreduced Signal-to-Noise Ratio (SNR) measurements, increased bitdetection errors, and related operational failures. The cause of thisdegradation may be the frequency synthesizer that converts thereference frequency to the unique master frequency required bya specific receiver. These degradations are generally the result ofresidual high-frequency phase noise introduced by tunable frequencysynthesizers. In this paper, we quantitatively model theeffects that reference frequency phase noise has on the receiver’smaster timing chain, specifically: L-band to baseband downconversionerrors and user clock errors. These timing chain errorsthen impact GPS receiver phase tracking, bit detection, and SNRestimation. Although a GPS receiver can compensate for "lowfrequency"noises by calculating the oscillator’s bias and drift rate,it cannot compensate for excess high-frequency phase noise.Using the analysis presented here, we can then understand whyhigh-frequency sideband performance or small time differenceAllan variance performance typical of a high-quality referencestandard is needed, and why care is needed in selecting a generalpurposesynthesizer. Measurements with a MITEL/Plessy GPSBuilder-2 kit and P(Y) Commercial off-the-Shelf (COTS) receiverdriven by an external oscillator will be presented to illustrate theanalysis. The main contributions of this work are to: (1) derive theindustry rule-of-thumb of maintaining –80 to –100-dBc sidebandsover the GPS front-end bandwidth and the oscillator’s bandwidthfh; (2) understand how the upper bandwidth of the oscillatornoise spectrum, fh, enters into the problem; and (3) translate oscillatorspecifications in Allan variance and other domains into thesideband domain that many circuit designers seem to prefer.Tetewsky, A.K.; Lozow, J.B.The Effects of Ground Planes on Rotating GPS Antennas55th Navigational Technology for the 21st Century: Institute of Navigation AnnualMeeting, Cambridge, MA, June 28-30, 1999, Proceedings (A00-18180 03-32),Alexandria, VA,1999, p. 289-300.Time-varying changes in the orientation of a single GlobalPositioning System (GPS) antenna with respect to the circularlypolarized GPS signals in space can produce carrier phase modulation.However, popular GPS references and software tools do notinclude these terms and typically show only the contribution oftranslational terms to the GPS group delay and carrier phase measurements.With the emerging Real-Time Kinematic (RTK) differentialpositioning and attitude fix algorithms requiring cm-levelaccuracy, coupled with arrays of antennas that are not constrainedto be coplanar, orientation effects must be modeled. Thispaper presents a general theory of the phase effects introducedby changing antenna orientation and extends previous work byallowing ground plane effects to be modeled. Although computermodels will ultimately be needed in order to account forground-plane effects, by working with some simplified coordinateframes and motion models, additional factoring of the polarizationfunctions into a pure orientation term plus a small residualspin modulation function yields valuable insights into the physicsand interpretation of the polarization terms. A brief review of thephase wrapup problem and the original derivation based on theHertzian (infinitesimal) dipole approximation without a groundplane is presented. Next, directivity patterns for a smallmicropatch antenna with ground planes are evaluated, and thephase wrapup and spin modulation polarization effects are calculated.Because the directivity pattern holds only for the upperhalf-space, care must be used when calculating element responsesto exclude contributions from beneath the ground plane. Dueto this additional complexity, only the results for the antenna’sboresight aligned with the spin axis will be covered. Although thetechnique is general, other geometries introduce significantamounts of algebra, and results are not yet available. The impactthat the polarization phase has on raw (pseudorange and carrierphase) measurements, navigation fixes, and general receiveroperation are also discussed.1999 Published Papers 89

Tetewsky, A.K.; Youngberg, J.W.A Users Perspective on the Continuing Evolution of GPS Simulators55th Institute of Navigation Annual Meeting, Cambridge, MA, June 28-30, 1999,Proceedings (A00-18180 03-32), Alexandria, VA, Institute of Navigation, 1999, pp. 581-596A Global Positioning System (GPS) simulator is a valuable piece oftest equipment. For the receiver developer, it provides a controllablesignal source. For the integrator, it places a receiver (and,often, other navigation sensor) on a movable host platform. Forthe analyst, it produces an environment to enable studying theoperational performance of a system. Over the last decade, wehave seen developers’ and testers’ expectations of simulatorcapability and technical performance increase in parallel to theevolution and maturing of end-user missions. This paper identifiesthe trends we have seen and projects their implications onfuture simulator hardware and software.Thompson, M.T.; Thornton, R.D.; Kondoleon, A. (reprinted)Flux-Canceling Electrodynamic Maglev Suspension: Part I Test Fixture Design andModelingIEEE Transactions on Magnetics, Vol. 35, No. 3, May 1999, pp. 1956-1963The design and analysis of a scale-model suspension test facilityfor magnetic levitation (maglev) is discussed. We describe techniquesfor the design, construction, and testing of a prototypeElectrodynamic Suspension (EDS) levitation system. The viabilityof future high-temperature superconducting magnet designs formaglev has been investigated with regard to their application toactive secondary suspensions. In order to test the viability of anew "flux-canceling" EDS suspension, a 1/5-scale suspensionmagnet and guideway was constructed.The suspension was testedusing a high-speed rotating test wheel facility with a linearperipheral speed of up to 84 m/s (300 km/h). A set of approximatedesign tools and scaling laws has been developed in order toevaluate forces and critical velocities in the suspension.Tingley, R.; Pahlavan, K.Propagation Measurement Using Antenna ArrayElectronics Letters, Vol. 35, No. 15, pp. 1211-12The design and construction of a 2.4-GHz antenna array suitablefor measurement of the time, angle, and complex amplitude ofpath arrivals in an indoor radio channel are described. Calibrationof the array is facilitated with the aid of an anechoic chamber. Anoptimal least-squares processor is derived, which compensatesfor systematic calibration errors. Early measurement results arepresented, and the future direction of the research is indicated.Toomey, K.; Seman, A.Enabling Technologies for Cost-Effective Shipboard Situational Awareness - Reduced ShipsCrew by Virtual Presence (RSVP) - 1999 Advanced Technology Demonstration (ATD)12th Ship Control Systems Symposium, The Hague, Netherlands. Sponsored by: SCSCurrent U.S. Navy technology development thrusts and shipdesigns are being driven by pressures to reduce the costs ofacquisition and cost of ownership (life-cycle costs). As identifiedin the 1995 NRAC Study on "Life-Cycle Cost Reduction" and reiteratedin a 1996 NRAC Summer Study on "Damage Control andMaintenance for Reduced Manning," a majority of the total cost ofownership of a ship is operation and support costs. Of these costs,manning is identified as the predominant cost driver. As stated inthe 1995 NRAC study, reducing manning is not straightforward,and "impacts the complex relationship of manpower requirementsfor operating, maintaining, supporting, fighting, and savingthe ship. A rational approach to reducing manning requires a systemsengineering approach with in-fleet demonstrations of proofof principle." To address this problem, the Navy is pursuingchanges in doctrine and insertion of cutting edge technologyaboard selected commissioned ships. Technology insertion willinclude advanced sensors, wireless networking, distributed monitoring,processing, and advanced reasoning capabilities. Currentsystems such as the Damage Control System (DCS), automatedMachinery Control System (MCS), and the Integrated ConditionAssessment System (ICAS) provide some level of this capability.However, the full level of automated monitoring and situationalawareness/assessment required to safely reduce manning doesnot exist in these systems today. Reliable, accurate, and timelyautomated ship system assessment and awareness is required tosupport ship operation in a reduced crew environment. TheOffice of Naval Research-funded RSVP approach and demonstratedtechnologies capture the life-cycle cost reduction objectivesand will form the basis of a ship-wide systems approach capableof providing situational awareness of ship's systems and compartmentsnecessary for ship operation in a reduced manningenvironment. This paper will explore the necessary system architecturetrade-offs of capability, cost, power consumption, reliability,and commercialization associated with the elements of theRSVP approach.Vytal, J.Shipboard EMI/EMC Test Report for the Reduced Ships-Crew by Virtual Presence (RSVP)Advanced Technology Demonstration (ATD)Journal Announcement: USGRDR0004Electromagnetic Interference/Electromagnetic Compatibility(EMI/EMC) testing was conducted on board the USS Normandy(CG-60), a Ticonderoga Class Aegis Cruiser, in early April 1999. Thetests were made to determine a typical electromagnetic operatingenvironment for the RSVP RF communications system and toperform propagation measurements in the proposed 2.4-GHz ISMband. The scope of the testing included measurements of theelectromagnetic environment from 10 kHz to 3 GHz in three differentspaces aboard the ship, and 2.4-GHz propagation measurementsin those spaces. The spaces chosen were Main EngineRoom 2, Auxiliary Machinery Room 1 and Engineering CrewQuarters. Of particular interest for the EM measurements was theband at 2.4 GHz and those surrounding 100 MHz and 10.7 MHz,the proposed first and second Intermediate Frequencies (IFs) forthe RSVP receiver. While the testing revealed no serious problems,it must be remembered that these measurements are only asnapshot in time aboard a single ship. Testing onboard differentships may reveal significantly different results.Weigold, J.W.; Juan, W.H.; Pang, S.W.; Borenstein, J.T. (reprinted)Characterization of Bending in Single Crystal Si Beams and ResonatorsJournal of Vacuum Science & Technology B, Vol. 17, No. 4, July-August 1999, pp. 1336-1340Optical interferometry has been applied to determine the displacementof p(++) Si beams. Clamped-clamped Si beams andcantilevered beams were fabricated with short and long B diffusionprocesses and characterized. Measurements of beam bendingfor released Si structures with length varying from 50 to 1000µm, width varying from 5 to 15 µm, and thickness varying from 6to 37 µm were obtained. By taking advantage of an etch-diffusionprocess, thicker beams can be fabricated that have less bendingdue to stress gradients. A 6.0-µm-thick cantilevered beam had adeflection of 11.2 µm due to stress gradients, while a 36.7-µmthickbeam had a deflection of only 0.3 µm. Beams fabricated901999 Published Papers

using a dissolved wafer process with a 12-h B diffusion werefound to bend the same amount as those fabricated with a 4-hdiffusion. This indicates that bending in doped Si beams not onlydepends on the gradients in the B concentrations, it could also berelated to the distribution of dislocations. Using the deep-etchshallow-diffusion process, resonating elements that are 20 µmlong, 4 µm wide, and 28 µm thick were found to be perfectly flatwithout any bending.Xu, B.M.; Ye, Y.H.; Cross, L.E.; Bernstein, J.J.; Miller, R. (reprinted)Dielectric Hysteresis under Transverse Electric Fields in Sol-Gel Lead Zirconate TitanateFilms Deposited on ZrO 2 Passivated SiliconIntegrated Ferroelectrics, 1999, No. 1-4, pp. 19-31Lead zirconate titanate (PZT) thick thin films with a Zr/Ti ratio of52/48 have been prepared on ZrO 2 -passivated, 4-in diameter siliconsubstrates with a thermally-grown SiO 2 layer. Both the ZrO 2 -passivated layer and PZT film are deposited through a similarsol-gel process that can be used to make PZT films with thicknessesup to 5 µm. Using interdigitated electrode arrays on theupper surface, the dielectric and ferroelectric properties of PZTfilms are characterized, with emphasis on 1-µm thick films.Dielectric constant of over 1000 with dielectric loss of about 0.01is achieved for the films. Excellent symmetric hysteresis loops arealso obtained with the apparent remanent polarization of around20 µC/cm 2 and coercive field of 20 to 30 kV/cm. The resultsdemonstrate that the properties of these PZT films on ZrO 2 -passivatedsilicon substrates are comparable to that of the PZT films onPt-buffered silicon substrates, and they can be used to fabricatemicromachined, d(33)-mode unimorph bending transducers thatare expected to have much better performance than the conventional,d(31)-mode bending transducers.Xu, B.M. (reprint); Polcawich, R.G.; McKinstry, S.; Ye, Y.H.; Cross, L.E.; Bernstein,J.J.; Miller, R.Sensing Characteristics of In-Plane Polarized Lead Zirconate Titanate Thin FilmsApplied Physics Letters, Vol. 75, No. 26, December 27, 1999, pp. 4180-4182The sensing characteristics of in-plane polarized lead zirconatetitanate (PZT) thin films were studied and compared with thethrough-thickness polarized PZT films. The in-plane polarized PZTfilms were deposited on ZrO 2 -passivated silicon substrates andhad interdigitated electrode systems on the top surface; hence,they can be polarized in the film plane. This in-plane polarizationconfiguration separates the electrode spacing and film thicknessas independent variables; thus, the voltage sensitivity can beincreased by using wider electrode spacing even for fixed filmthickness. The results show that for films with a thickness of 1 µm,the voltage sensitivity of in-plane polarized PZT films can be morethan 20 times higher than that of the conventional, throughthickness-polarizedPZT films that were deposited on Pt-bufferedsilicon substrates.Zarchan, P.; Gratt, H.Adaptive Radome Compensation Using DitherJournal of Guidance, Control, and Dynamics, Vol. 22, No. 1, January-February 1999, pp.51-57A technique is presented for estimating a radar homing missile’sradome slope by using a nondestructive dither signal on theacceleration command. A planar example is presented in detailshowing how bandpass filtering is used to extract the radomeslope estimate and then to compensate for unwanted radomeaberration angle effects. A second example is presented showinghow Kalman filtering techniques can also be used for the sameplanar example to estimate the radome slope. Although theKalman filter approach does not yield superior radome slope estimates,it does provide a solid framework so that the radome slopeestimation technique can be extended to the more realistic threedimensionalcase where cross-plane slopes are important.Zimpfer, D.J.On-Orbit Flight Control Design for Kistler K-1 Reusable Launch VehicleGuidance, Navigation, and Control Conference, Portland, OR. Sponsored by: AIAA, pp.1266-1274This paper describes the on-orbit control design for the Kistler K-1Reusable Launch Vehicle Orbiting Vehicle (OV) stage. To meet theK-1 design goals, the control must provide fully autonomousoperation with minimal preflight reconfiguration. The on-orbitguidance, navigation, and control targets, guides, and controls thevehicle to and from orbit, maintains orientation to adequatelydeploy payloads, separates from the payloads to avoid collision,and provides spin-stabilization prior to long periods of orbital"sleep" operations. For control, the OV relies on a gimbaled OrbitalManeuvering System (OMS) and cold-gas thrusters. The flightcontrol algorithms are derived from the U.S. Space Shuttle andRussian Mir Space Station Control algorithms. This paperoverviews the on-orbit mission operations for the K-1 vehicle, presentsthe control algorithms developed, and summarizes predictedon-orbit control performance.1999 Published Papers 91

"Draper cultivates its employees’ ability to create innovative solutions to difficult technical problems. Pursuing patent protection for its employees’ inventiveideas enables Draper both to pursue its strategic mission and to recognize its employees’ valuable contributions to advancing the state-of-the-art intheir technical areas."William Elias,Draper’s Legal CounselDraper Laboratory, which was formed to contribute to scientific research and technological development,has a tradition of technical creativity. The disclosure of inventions is an important step in documentingDraper personnel’s creative efforts, and is a requirement under Laboratory contracts (and by an agreementwith Draper that all employees sign). Draper has an established patent policy and understands the value ofpatents in directing attention to effective individual accomplishments in science and engineering.Millions of U.S. patents have been issued since the first patent in 1836. Through December 31, 1999, 976Draper patent disclosures have been submitted to the Patent Committee since 1973; 484 of which wereapproved by Draper’s Patent Committee for further patent action. As of December 31, a total of 393 patentshave been granted for inventions made by Draper personnel. Eleven patents were issued for calendar year1999.On average, Draper’s Patent Committee typically recommends seeking patent protection for 50 percent ofthe disclosures received. "Draper’s high percentage of authorizations for patent action stems from the highquality of invention disclosures submitted by its employees, and from the seriousness with which Draperapproaches its goal of transferring technology," Elias stated. "Through patents, Draper can foster the use ofits innovative technologies in the commercial sector, as well as in the governmental arena. This benefitsDraper’s sponsors, its employees, and Draper itself."This year’s featured patent is: Autonomous Local Vertical Determination Apparatus and Methods for aBallistic Body.The following pages contain an overview of the technology covered in the patent, followed by the officialpatent abstract issued by the U.S. Patent Office.921999 Patents Introduction-- ---- -

Donald E. GustafsonDavid J. LuciaAutonomous Local Determination Apparatusand Methods for a Ballistic BodyDraper is at the forefront of the development of a new class ofGuidance, Navigation, and Control (GN&C) systems suitable foruse in gun-fired munitions, particularly artillery shells. Significantreductions in size, cost, and weight have been made possible, primarilyas a result of evolving Microelectromechanical Sensor(MEMS) technology for both gyros and accelerometers.Accurate local vertical (roll angle) estimates are required for guidedshells in order to correctly implement guidance commands. Acommon navigation system architecture for guided shells usesboth MEMS and Global Positioning System (GPS) data that aremerged in an integration filter. Although GPS data will typicallybe merged in a navigation filter to estimate local vertical as wellas position and velocity, it is desirable to have the capability ofestimating local vertical in the absence of GPS for several reasons:(1) GPS data will not be available early in flight before receiverreacquisition has been completed.(2) GPS data could drop out or be corrupted during flight.(3) Without proper initialization, the quaternion algorithms usedfor attitude angle calculations in the integration filter may notconverge properly due to nonlinearities in the algorithms.With no GPS data, the GN&C system is in an autonomous mode,and possible guidance functions include:(1) Implemention of pre-stored trajectory maneuvers.(2) Completion of guided maneuvers interrupted by GPS datadropouts.(3) In-flight compensation of path deflections due to atmosphericdisturbances (wind, density).In another scenario, the shell could be tracked externally andcommand guided to the target using only onboard MEMS sensorsfor navigation. Accurate roll angle estimates would be required toachieve satisfactory target miss distances at impact.It is not possible to estimate roll angle immediately after launch ofa shell due to the extremely large launch impulses (tens of thousandof g) and high angular rates (several hundred Hz) encountered.Current gyro technology does not allow roll rates to bemeasured under these conditions.This patent solves the problem of roll angle estimation during theflight of a ballistic shell when no prior roll angle information isavailable and only onboard inertial sensors are available.The onlysensors required are three strapdown gyros configured to allowangular rate measurements along all three body axes. Theapproach is based on the observation that the rate of change offlight path angle has a highly predictable component due to theaction of gravity. In the ideal case of a shell flying ballistically overa nonspinning Earth with zero spin rate and zero angle-of-attack,the flight path angle rate is a known function of flight path angle,speed, and gravitational acceleration. If no navigation errors arepresent, then flight path angle rate in local vertical coordinates isknown precisely and body pitch and yaw rates are known functionsof roll angle. This allows roll angle to be calculated from themeasured pitch and yaw rates.In reality, measured body rates are corrupted by sensor errors, andangle-of-attack and sideslip are nonzero with significant unpredictablecomponents. The problem is compounded by the factthat the gravity-induced "signal" power is generally much lowerthan the "noise" power due to measurement errors and disturbances.In order to obtain useful local vertical estimates, anextended Kalman filter is employed that significantly reduces theeffects of the noise and provides near-optimal roll angle estimates.The measurements (body rates) are nonlinear functions ofroll angle. Filter performance is optimized by using a set of rectangularstates rather than a roll angle state. With this state definition,the measurements become linear in the states, and theinherent nonlinearity in the problem is transferred to the filterpropagation step, in which the driving noise has state-dependentterms that are handled using statistical averaging methods. Thistransformation significantly reduces filter convergence problemsat low signal/noise ratios and achieves consistently lower roll rateerrors than filters that employ nonlinear measurement models.The problem of roll rate estimation in gun-fired ballistic shells hasapparently not been considered before, since effective means ofmeasuring body rates have not been available before the adventof micromechanical sensors.This invention was demonstrated successfully during flight testsconducted as part of the Extended-Range Guided Munition(ERGM) Demonstration Program recently completed by Draper asthe prime contractor. The flight tests were conducted using a 5-inNavy gun-launched projectile with a launch impulse of 6,800 g.Autonomous Local Vertical Determination Apparatus and Methods for a Ballistic Body 93

United States Patent (1 ~JGUSlar!illn t l al.111111 1111111(II) "'alt nt N um ~r : 5,886,257(4S] Ualt or I'alt nl:Mar. 2J, 19991$-11 ."\JTONO~IOlIS WCJ\L \'Ul'IlCo\Lnt: I't .HMI NArU IN 41,'4KAnss MWMt: UIOIIS ~ OK'" IIAI .I .I"TII: l~mY1751 [nV1t>gy IMoIrxt).8< ............. (199.') ICC C""~ "" '.II;~'M'';'~",.. Fwu..-!· (AbMrO11>< .... lOT'" .'" IIIOODI«i 10 • bollhlk bool)' Iu pnMo.Io.D .uioWGl

onald E. GustafsonbiographiesDonald E. Gustafson is a Principal Member of the Technical Staff in the Guidanceand Navigation Division at Draper. He has over 30 years experience in the conceptualdesign and analysis of GN&C systems. While at Draper, he has workedprimarily on the design and analysis of integrated INS/GPS navigation systemsand is currently working on developing and testing new navigation algorithmsthat offer high immunity to jamming and interference. He has also worked on robotic vision andunderground navigation systems for mining applications. Previously, he was at MIT LincolnLaboratory, where he worked on target surveillance using antenna arrays. Prior to this, he wascofounder and Vice President of Scientific Systems for 11 years, where he worked in the areas ofaircraft failure detection, adaptive control of plastic injection molding machines, biomedical signalprocessing, meteorological satellite data processing, and financial forecasting. From 1966 to1977, he was with the MIT Instrumentation Laboratory (later Draper), where he worked primarilyon Apollo navigation system design and computerized electrocardiogram interpretation. Hereceived a BS in Electrical Engineering from Akron University in 1962, an MS in ElectricalEngineering from Santa Clara University in 1966, and a PhD in Instrumentation and AutomaticControl from MIT in 1973. Dr. Gustafson was the co-recipient of the 1995 Draper Best TechnicalPublication Award for "A Micromechanical INS/GPS System for Guided Projectiles." He is coholderof a U.S patent for "First-Order Phase-Lock Loop," and has authored more than 30 technicalpapers.avid J. LuciabiographiesDavid J. Lucia, Major USAF, received a BSc (high honors) in Aerospace Engineering from theUniversity of Texas (1988) and an MS in Aeronautical and Astronautical Engineering from MIT(1995). On receiving his commission as an Air Force Officer, he spent 4 years assigned to theMILSATCOM Terminal Programs Office at Hanscom AFB, MA, where he worked as a systemsengineer and project manager on the MILSTAR satellite communications system. He next spent2 years at Draper as a research assistant in GN&C while he pursued his Master’s degree at MIT.On graduation, he was assigned to the Space Warfare Center in Colorado Springs, CO, where hewas a space systems analyst specializing in operational modeling and analytical support of theGPS space segment for Air Force Space Command. He is currently attending the Air ForceInstitute of Technology working toward his PhD in Aeronautical Engineering. His currentresearch activities include reduced-order computational fluid dynamics, control/fluids/structuresinteractions, and design environment development for high-speed flow fields.~---- -Autonomous Local Vertical Determination Apparatus and Methods for a Ballistic Body: Biographies 95

In addition to the featured patent, the following pages contain a list of all other patents issuedby the U.S. Patent Office to Draper engineers during the 1999 calendar year.Bernstein, JonathanMethod for Fabricating a Tunneling Sensor with Linear Force RebalancePatent No. 5991990, Issued 30 November 1999A tunneling sensor has a pair of force rebalance capacitors thatare used in a push-pull relationship to provide a rebalance forcethat is a linear function of applied rebalance voltages, which leadsto an output voltage that is linearly related to input acceleration.The tunneling sensor comprises a plate electrode that is formedfrom and attached to a silicon substrate by a pair of torsional flexures,which provide an axis of rotation for the plate electrode. Apendulous mass is formed on a first end of the plate electrode,and a tunnel-effect contact is formed on a second end of the plateelectrode. A pair of torque rebalance bridge electrodes areformed on the substrate to span the plate electrode. A tunneleffecttip is formed on the substrate to be proximate to the tunnel-effectcontact and in line with the rotational path that thetunnel-effect contact takes when the plate electrode is rotated.Bernstein, Jonathan; Cunningham, BrianElectrostatic Memory Micromirror Display SystemPatent No. 5903383, Issued 11 May 1999(1) An electrostatic memory micromirror display system, includingan array of electrostatic memory and display assemblies eachincluding a memory and display element having a mirror surfaceand a conductive medium; (2) a support device for movablymounting the memory and display element; and a set of electrodes,including at least first and second electrodes proximate tothe conductive medium; (3) means for selectively applying a voltageto the first electrode of predetermined ones of the memoryand display assemblies to establish an electrostatic field betweenthe first electrode and the conductive medium to urge the associateddisplay element from a first position toward a second positionand for applying a voltage to the second electrode of theassemblies to establish an electric field between the second electrodeand the conductive medium to set the memory and displayelement in the second position is disclosed.Bernstein, JonathanMonolithic Micromachined Piezoelectric Acoustic Transducer and Transducer Array andMethod of Making SamePatent No. 5956292, Issued 21 September 1999(1) A monolithic micromechanical piezoelectric acoustic transducerwith integrated control circuit includes a support member;(2) a piezoelectric medium disposed on the support member; (3)first and second electrodes engaging the piezoelectric medium;(4) a control circuit monolithically integrated with the piezoelectricmedium and electrodes on the support member and includinga switching circuit for selectively interconnecting theelectrodes with an I/O bus and a signal processing circuit for conditioningsignals propagating between the electrodes and the I/Obus; (5) an array of such acoustic transducers that form an acousticretina; (6) a method of making such transducers and arrays.Eyles, DonaldSystem and Method for Automatically Executing Decisional RulesPatent No. 5978786, Issued 2 November 1999A system and method for automatically evaluating a decisionalrule containing a task and a condition that must be fulfilled beforethe task can be performed and for automatically performing thetask when or whenever the condition is fulfilled in which a decisionalrule is entered into computing means, parsed, and thenprocessed such that automatic and continuing iterative evaluationsof whether the condition is fulfilled are provided until thecondition is fulfilled and then the task associated with the decisionalrule is automatically performed and whereby further processingis resumed only after the condition is fulfilled.Greenspan, RichardRadio-Wave Reception System Using Inertial Data in the Receiver BeamformingOperationPatent No. 5917446, Issued 29 June 1999A receiver system on a mobile host vehicle platform includes aninertial sensor embedded in an antenna ground plane that supportsan array of antenna elements. The beamformer within thereceiver system determines the beamforming weights by incorporatinginertially-generated signals representative of the attitudeof the receiver system and location data identifying thelocation of Global Positioning System (GPS) satellites. As the hostvehicle moves, the beamformer generates the appropriate gainpattern based on the inertial data of the current attitude and theGPS location data. The beamformer, in particular, performs a spatialfiltering function that is characterized by high-gain profiles inthe direction of transmission of selected ones of the GPS terminals,thereby effectively suppressing signals originating from jammersand other sources of Radio Frequency Interference (RFI).961999 Patents Issued

Greiff, Paul; Brezinski, PaulGettering Enclosure for a Semiconductor DevicePatent No. 5929515, Issued 27 July 1999Gettering enclosures for semiconductor packages, comprising anenclosure having a cavity for accommodating a semiconductordevice, is disclosed; a gettering chamber (disposed above thesemiconductor device) communicating with the cavity, comprisinga getter precursor secured to the cavity and spaced from awall of the cavity, wherein the wall is transparent to laser light toallow a beam of laser light to strike the getter precursor and sputtersame on the walls of the gettering chamber is disclosed.Greiff, PaulMicromechanical Accelerometer Having a Peripherally Suspended Proof MassPatent No. 5969250, Issued 19 October 1999A monolithic, micromechanical vibrating beam accelerometerwith a trimmable resonant frequency is fabricated from a siliconsubstrate that has been selectively etched to provide a resonantstructure suspended over an etched pit. The resonant structurecomprises an acceleration-sensitive mass and at least two flexibleelements having resonant frequencies. Each of the flexible elementsis disposed generally collinear with at least one acceleration-sensitiveaxis of the accelerometer. One end of at least one ofthe flexible elements is attached to a tension relief beam for providingstress relief of tensile forces created during the fabricationprocess. Mass support beams having a high aspect ratio supportthe mass over the etched pit while allowing the mass to movefreely in the direction collinear with the flexible elements. Alsodisclosed is a method for fabricating such an accelerometer.Further disclosed is an alternative embodiment of the aforementionedaccelerometer characterized by a low profile, and alternativeplanar processing methods for fabricating these and otherembodiments.Ward, Paul; Hildebrant, Eric; Niles, Lance; Weinberg, Marc; Kourepenis,AnthonySplit Electrode to Minimize Charge Transients, Motor Amplitude Mismatch Errors, andSensitivity to Vertical Translation in Tuning-Fork Gyros and Other DevicesPatent No. 5911156, Issued 8 June 1999A micromechanical tuning-fork gyroscope having two centerelectrodes is disclosed. The two center electrodes are excitedwith bias potentials of opposite polarity. The oppositely biasedcenter electrodes provide electrical symmetry across the gyroscopeand thereby reduce charge transients and sensitivity to verticaltranslation. Currents injected directly into the proof massesare equal and opposite and thus cancel. Motor lift forces actingon the proof masses and interleaved electrodes are equal, andhence the proof masses move in pure translation, thereby reducingin-phase bias. Further, any pure translation normal to theplane of the gyroscope does effect sense axis output signals.Weinberg Marc; Pinson, JohnGuard Bands Which Control Out-of-Plane Sensitivities in Tuning-Fork Gyroscopes andOther SensorsPatent No. 5892153 Issued 6 AprilGuard bands that reduce or eliminate force and sensitivity associatedwith proof mass motion normal to the substrate as a result ofvoltage transients are disclosed. The guard bands are biased toreduce the coupling ratio to zero and thereby enable starting andimproved performance. Various configurations of guard bandsmay be employed, including distinct inner and outer guard bands,distinct inner guard bands only, extended sense electrodes belowinner sensing combs with no outer guards, distinct outer guardbands with extended sense electrodes, and sense electrodesextended beneath both drive and sensing combs.Weinberg, Marc; Niles, Lance; Ward, Paul; Hildebrant, Erik; Kourepenis,Anthony; Hopkins, Ralph; Cho, StevenTrenches to Reduce Charging Effects and to Control Out-of-Plane Sensitivities in Tuning-Fork Gyroscopes and Other DevicesPatent No. 5952574, Issued 14 September 1999Trenches that reduce or eliminate force and sensitivity associatedwith proof mass motion normal to the substrate as a result ofvoltage transients are disclosed. The trenches provide increasedseparation between interleaved comb electrodes and the substrate,and thereby also reduce the comb lift to drive ratio. Thetrenches are typically formed directly below the interleaved combelectrodes, but may also be formed below other suspended portions.Trench depth is from 6 to 10 µm and provides a comb electrodeto substrate separation of approximately 8.5 to 12.5 µm.1999 Patents Issued 97

The Charles Stark Draper Prize was established in 1988 to honor the memory of Dr. Charles Stark Draper,"the father of inertial navigation." The Prize was instituted by the National Academy of Engineering andendowed by Draper Laboratory, and is recognized as one of the world’s preeminent awards for engineeringachievement. It honors individuals, who like Dr. Draper, developed a unique concept and put it into practicein ways that contributed significantly to the advancement of science and technology, as well as thewelfare and freedom of society. Once awarded biennially, the Draper Prize is now awarded annually.THE NOMINATION PROCESSNominations of candidates for the Draper Prize, awarded toliving persons from any country, are sought from membersand foreign associates of the U.S. National Academy ofEngineering, National Academy of Science, and Institute ofMedicine; members and foreign associates of academies ofengineering worldwide; members of recognized U.S. andinternational societies; and other individuals deemed eligibleby the National Academy of Engineering who represent abroad spectrum of engineering disciplines.For more information on the nomination process, contact thePublic Affairs Office at the National Academy of Engineeringat (202) 334-1237.THE 1999 PRIZEThe 1999 Draper Prize was presented to Drs. Charles Kao,Robert Maurer, and John MacChesney on February 22, 2000,during a ceremony at the Department of State for theirachievements in spearheading advances in fiber-optic technology.The importance of fiber optics and its impact on informationtechnology and worldwide telecommunicationscannot be overstated. The pioneering efforts of these threescientists have revolutionized the telecommunications industry;communications as we now know it, including theInternet, videoconferencing, electronic commerce, and highqualitylong-distance telephone service, would not exist withouttheir ground-breaking work.Draper President Vince Vitto stated, "It is fitting that the menwho developed the technology enabling the creation of fiberoptics are awarded the Draper Prize. Fiber technology hasallowed for worldwide information transport and has had aprofound effect on the global information infrastructure."DOCTOR CHARLES KAOWhile working at ITT’s Standard Telecommunications Laboratories in the 1960s, Dr. Kaoconsidered using optical fiber for communication, instead of the bulky copper wire usedat the time, and was the first to propose publicly a practical application for fiber-optictelecommunication. Analyzing samples from fiber makers, he concluded that fiber signalloss resulted from impurities and that silica glass could achieve the performance neededfor successful communication. His concept for long-distance communication over singlemodefibers, developed with colleague George Hockham, was published in 1966. Dr. Kao’sanalytical basis for the development of optical fibers for telecommunications inspired theinterest of other researchers world wide, and stimulated further research and developmenton glass fiber waveguides for communications.Dr. Kao received a BSc in 1957 and a PhD in 1965, both in Electrical Engineering, from theUniversity of London. He is currently Chairman and Chief Executive Officer of TranstechServices Ltd.100The Charles Stark Draper Prize

DOCTOR ROBERT D. MAURERDr. Maurer led a team of researchers at Corning Inc. (including co-inventors Donald Keckand Peter Shultz) who designed and produced the first low-loss optical fiber in 1970.While most researchers at the time were investigating compound glass, Mauer concludedthat fused silica was more promising. There were some difficulties to overcome, however.It was expected that fabricating fiber from silica, with its high melting point, would beimpractical due to imperfections introduced in the manufacturing process. Maurer and hiscolleagues solved some of these problems by devising a fiber preform from the vaporphase on a mandrel. When the mandrel was removed, the glass collapsed, and was drawninto a fiber. The material was then chemically graded to provide the low index of refractionand the higher index core needed for wave guiding. This materials design is in usetoday in all optical fiber for long-distance communication.Dr. Maurer received BS and PhD degrees in Physics from the University of Arkansas,Fayetteville (1948) and the Massachusetts Institute of Technology (1951), respectively.DOCTOR JOHN MACCHESNEYIn 1974, Dr. MacChesney, then of Bell Laboratories, announced a process for the controllableand reproducible manufacture of low-loss optical fibers. His process, ModifiedChemical Vapor Deposition (MCVD) solved problems of purity and contamination fromtrace water, allowing the complex doping profiles needed for the optical waveguides tobe achieved. The MCVD technique was introduced worldwide, and enabled the timely dispositionof optical fiber. AT&T installed the first optical fiber communication systembetween Washington, D.C. and Boston, MA in 1981.Dr. MacChesney received a BA from Bowdoin College and a PhD in Geochemistry fromPennsylvania State University. He is currently a Research Fellow at Lucent Technologies.PREVIOUS RECIPIENTS1997: Vladimir Haensel for the development of the chemical engineering process of"Platforming" (short for Platinum Reforming), which was a platinum-based catalystto efficiently convert petroleum into high-performance, cleaner-burning fuel.1995: John R. Pierce and Harold A. Rosen for their development of communication satellitetechnology.1993: John Backus for his development of FORTRAN, the first widely used, general purpose,high-level computer language.1991: Sir Frank Whittle and Hans J. P. von Ohain for their independent development of theturbojet engine.1989: Jack S. Kilby and Robert N. Noyce for their independent development of the monolithicintegrated circuit.The Charles Stark Draper Prize 101

The Draper Distinguished Performance Awards for Fiscal Year 1999 were presented totwo teams by Chairman of the Board Dr. Robert Hermann and President and CEO VinceVitto at the Annual Members Dinner on October 6, 1999.Established in 1989, the Draper Distinguished Performance Awards recognize outstanding technical achievements by teams orindividuals during the previous fiscal year. These achievements must represent a high standard of excellence, provide significantbenefit to the Laboratory, and be considered a major advance by the outside community.SMALL AERIAL RECONNAISSANCE DEMONSTRAITON (SARD) PROJECTMichael Piedmonte, Paul DeBitetto, Tony Lorusso, and ChrisSanders were recognized for their efforts on the SARD project.The team produced a rotorcraft that features small electronicspackaging and a simple, user-friendly human/computer interface.The SARD also demonstrated autonomous landing andtrue autonomous flight.(from left to right)Paul DeBitetto, Chris Sanders,Michael Piedmonte and Tony Larusso98The Draper Distinguished Performance Award~-- ---- -

All Draper employees (excluding Officers) are eligible, including both full-time and part-time employees and individuals who haverecently left the Laboratory. Any Draper employee may nominate an individual or team for consideration. Nominations arereviewed by the Distinguished Performance Award Committee, which selects recipients and forwards their recommendations tosenior management for approval.DEVELOPMENT OF CONTROL DESIGNS FOR THE INTERNATIONAL SPACE STATION (ISS) ASSEMBLY SHUTTLE FLIGHT STS-88Gene Rosch, Kim Kirchwey, Michael Martin, and Rob Hall receiveda team award for their efforts in developing the control designsfor the ISS Assembly Shuttle Flight STS-88. These designs werekey to the success of the Shuttle-assisted assembly operations.The mated configuration of the Shuttle and the two SpaceStation modules was unique, and one of the modules was thelargest payload ever handled by the Shuttle arm. The team hasalso won several awards from the Johnson Space Center for theirwork on this effort.(from left to right)Gene Rosch and Kim Kirchwey[missing from photo:Michael Martin and Rob Hall]~--- -The Draper Distinguished Performance Award 99

"To perform and contribute to the support and advancement of scientific research, technology,and development, and to initiate, maintain, and engage in educational activities in the sciencesand allied subjects."This is a key tenet of Draper’s mission, rooted in its articles of incorporation. Draper Laboratory has strongly promoted and supportedadvanced technical education through numerous in-house, cooperative, and research programs with universities and collegesthroughout the country. The Laboratory actively recruits graduate and undergraduate students to work on real, technicalproblems. Students have the opportunity to work on challenging projects from conception through development and implementation.The Education Office administers such programs as:The Draper Fellow ProgramThe Undergraduate Cooperative ProgramThe Undergraduate Research Opportunities ProgramMIT-Draper Technology Development PartnershipThe Professional Summer ProgramProfessional Development TrainingThe Gordon Institute ProgramThe University Research ProgramConferences and Technical SymposiaPublicationsGovernment Residents, Interns, and Associates ProgramsIn 1999, the 28th Professional Summer Program enrollment exceeded 150 participants. Over 4600 students have attended thesesessions since inception. Also in 1999, Draper, along with the AIAA, the IEEE, and the Russian Academy of Navigation and MotionControl cosponsored the 6th Saint Petersburg International Conference on Integrated Navigation Systems in May. Nearly 200attendees from 60 organizations in 11 countries participated. In addition, the Draper-Sponsored University Research Program continuedwith nearly $2,000,000 of Draper-funded efforts at various universities.During 1999, the Draper Fellow program consisted of approximately 60 students from MIT and several other universities. Abstractsof graduate theses completed this year are available on the Draper Website at draper.com. The list of completed theses follows:Bibeau, R.T.; Supervisors: Rubenstein, D.; Ramnath, R.V.; Peraire, J.Trajectory Optimization for a Fixed Trim Reentry Vehicle Using Direct Collocation withNonlinear ProgrammingMaster of Science Thesis, May 1999Billingsley, G.O.; Supervisors: Jacobson, S.W.; Kuchar, J.K.; Peraire, J.Head-Up Display Symbology for Ground Collision AvoidanceMaster of Science Thesis, May 1999Brown, P.D.; Supervisors: McConley, M.W.; Deyst, J.J.; Peraire, J.Micro Air Vehicle Control Design: A Comparison of Classical and Dynamic InversionTechniquesMaster of Science Thesis, June 1999Cerminaro, A.M.; Supervisors: Kaiser, K.W.; Nelson, F.C.; Fermental, D.W.Simulation of Internal Damping in a Rotating System Supported by Magnetic BearingsMaster of Science Thesis, April 1999102Education Activities at Draper Laboratory-- ---- -

Chauddhry, A.I.; Supervisors: Kang, D.S.; Kuchar, J.K.; Peraire, J.Navigation of a High-Velocity Tele-Operated Ground Vehicle through an Obstacle-RichEnvironmentMaster of Science Thesis, September 1999Gerrish, N.D.; Supervisors: Borenstein, J.T.; Fitzgerald, E.A.; Hobbs, L.W.A Miniaturized Single-Crystal Silicon Solar Cell Array for a MEMS Power SourceMaster of Science Thesis, June 1999Hall, W.D.; Supervisors: Odoni, A.R.; Adams, M.; Barnhart, C.; Feron, E.; Gilbo, E.;Kolitz, S.Efficient Capacity Allocation in a Collaborative Air Transportation SystemDoctor of Philosophy Thesis, May 1999Hsieh, P.Y.; Supervisors: Kumar, K.; Reif, R.; Fitzgerald, E.A.DC Magnetron Reactive Sputtering of Low Stress AlN Piezoelectric Thin Films for MEMSApplicationMaster of Science Thesis, February 1999Jamoom, M.B.; Supervisors: McConley, M.W.; Feron, E.; Peraire, J.Constrained Optimization for Hierarchical Control System DesignMaster of Science Thesis, June 1999Jenkins, S.N.; Supervisors: Appleby, B.; Spearing, S.M.; Peraire, J.Investigation of Curved Composite Panels Under High-g LoadingMaster of Science Thesis, June 1999Johnson, H.D.; Supervisors: Lanzilotta, E.J.; Rowell, D.; Sonin, A.A.Real-Time Model Identification for Ground Vehicle Trajectory Estimation Using ExtendedKalman Filter Residual AnalysisMaster of Science Thesis, June 1999Kaliardos, W.N.; Supervisors: Hansman, R.J.; Kang, D.S.; Weiss, S.I.; Feron, E.;Pratt, G.A.; Peraire, J.Semi-Structured Decision Processes – Conceptual Framework for Understanding Human-Automation SystemsDoctor of Philosophy Thesis, September 1999Korka, D.A.; Supervisors: Anderson, J.; Verghese, G.; Smith, A.C.Kalman Filtering for an Aided Inertial Navigation SystemMaster of Science Thesis, May 1999Kwok, P.; Supervisors: Breuer, K.S.; Weinberg, M.S.; Qiu, T.; Sonin, A.A.Fluid Effects in Vibrating Micromachined StructureMaster of Science Thesis, June 1999Kyemba, H.R.; Supervisor: Lentz, K.P.Analysis of the Applicability of a General-Purpose Operating System for Time-CriticalTasksMaster of Science Thesis, May 1999Lad, P.B.; Supervisors: Wender, P.; Sharon, A.; Sonin, A.A.Design and Evaluation of an Automated Fiber-Optic Untwisting MachineMaster of Science Thesis, June 1999Lynch, M.A.; Supervisors: Haus, H.A.; Kelleher, W.; Smith, A.C.Orthogonal Polarization Fiber-Optic Gyroscope with Improved Bias DriftMaster of Science Thesis, May 1999Neelon, J.; Supervisors: Proulx, R.J.; Cefola, P.J.Orbit Determination for Medium Altitude Eccentric Orbits Using GPS MeasurementsMaster of Science Thesis, February 1999Nuzzo, N.; Supervisors: D’Souza, C.N.; Schmidt, G.T.Effects of Propagation Techniques on Relative GPS NavigationMaster of Science Thesis, February 1999Perel, R.Y.; Supervisors: Mangoubi, R.S.; Annaswamy, A.M.; Peraire, J.Learning Control for a Class of Discrete-Time, Nonlinear SystemsMaster of Science Thesis, June 1999Peskin, J.T.; Supervisors: Anderson, J.; Slotine, J.J.; Smith, A.C.Force Control of Electro-Hydraulic Actuators in an Underwater Fish-Like VehicleBachelor and Master of Science Thesis, May 1999Phan, L.N.; Supervisors: DeBitetto, P.A.; Sarma, S.E.; Sonin, A.A.Collision Avoidance via Laser RangefindingBachelor and Master of Science Thesis, May 1999Rogers, K.E.; Supervisors: Vander Velde, W.E.; Mangoubi, R.S.; Tsitiklis, J.N.;Feron, E.; Peraire, J.Scheduling of Costly Measurements for State Estimation Using Reinforcement LearningDoctor of Philosophy Thesis, September 1999Sammon, R.; Supervisors: DeBitetto, P.A.; Teller, S.; Smith, A.C.A Mapping System for an Autonomous HelicopterMaster of Science Thesis, June 1999Shenck, N.S.; Supervisors: Rosenstrach, P.A.; Paradiso, J.A.; Smith, A.C.A Demonstration of Useful Electric Energy Generation from Piezoceramics in a ShoeMaster of Science Thesis, May 1999Smith, J.E.; Supervisors: Proulx, R.J.; Cefola, P.J.; Peraire, J.Application of Optimization Techniques to the Design and Maintenance of SatelliteConstellationsMaster of Science Thesis, June 1999Tan-Hsin Fu, L.; Supervisors: Turkovich, J.J.; Pena-Mora, F.; Smith, A.C.Adaptation of Cairo Meeting Environment Toward Military Collaboration EffortsMaster of Science Thesis, May 1999Thele, J.D.; Supervisors: Kang, D.S.; Leonard, J.J.Control Interface for a High-Velocity Tele-Operated RobotMaster of Science Thesis, February 1999Vuong, H.F.; Supervisors: Abramson, M.R.; Feron, E.; Peraire, J.Modeling and Analysis of Software Specifications for an Autonomous Aerial VehicleMaster of Science Thesis, May 1999White, R.D.; Supervisors: Weinberg, M.S.; Feng, Z.; Sonin, A.A.The Effects of Mechanical Vibration and Impact on the Performance of a MicromachinedTuning-Fork GyroscopeMaster of Science Thesis, May 1999Zhao, W.; Supervisors: Magnanti, T.L.; Kolitz, S.; Williams, J.R.Multiple Autonomous Vehicle Mission Planning and ManagementMaster of Science Thesis, June 1999Education Activities at Draper Laboratory 103

104Education Activities at Draper Laboratory- -